]> Individual

pki@varwof.com https://varwof.com

Network Working Group This document defines the Capability Language Core (CLC), a minimal, executable language for describing what an agent is authorized to do. It defines the capability identifier grammar, the entailment relation between a grant and an operation, intersection of grants from multiple sources, the constraint model, and a deterministic decision function with stable reason codes and a three-valued verdict (allow, deny, allow_unresolved). The language is carrier-neutral: it defines what is evaluated, not how it is carried or trusted. Trust models, native verification, execution lifecycle, and receipt or token formats are out of scope (Section 11). Conformance is exercised by a published corpus of 154 vectors and 1184 property cases; three implementations (Go, Python, TypeScript) that share an author pass both. Implementation conformance and this document's claim of a conformance class are separate. An implementation conforms to CLC-A when it meets the obligations Section 12 lists for that class, and it may claim that conformance on its own, whatever other implementations exist. Section 12 additionally sets a maturity bar for this document's claim — two independent implementations agreeing on verdict and reason. That bar is not met here: as Section 12 states under "Independence of implementations", the three implementations named in the README share an author, so their agreement is a regression test for the text, not independent validation. CLC-A is still claimed by this revision as the baseline authorization class, whose obligations are implementable and exercised by a published corpus; the independent-implementation threshold is recorded as unmet. The evidence-side class CLC-E is not claimed: its relations, value grammar, reference implementation and corpus ship here. This revision also folds the delegation containment relation into the document (Section 13): Contains(parent, child) decides whether a child grant stays inside a parent's declared boundary — the single question a delegation chain asks at every hop that the core's entailment and intersection do not answer. It ships with the conformance class CLC-D and its own corpus, is strictly additive, and changes no CLC-A verdict, reason code, or vector. It absorbs the previously separate experimental containment extension, which is retired.

Introduction CLC-v1 is a universal minimal language that can be consistently presented on either the authorization side or the evidence side. The two sides share five foundational abstractions: Foundation Authorization Side Evidence Side Action Operation (concrete request) ObservedAction (asserted effect) Identity CapabilityId (class-level) ActionId (instance-level) Binding Entailment (grant ⊆ operation) Match (evidence ↔ action) Constraint Grant params / limits Evidence requirements / freshness Verdict allow / deny / allow_unresolved SATISFIED / UNSATISFIED Conventions. The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here. "UTC instant" is an timestamp. "JCS" is the JSON Canonicalization Scheme ; "I-JSON" is . The language structure is identical on both sides; only the direction differs: Authorization: "I grant you permission to do X" Evidence: "I have evidence that X was done" The design principles behind this core — including what the language deliberately refuses (no control flow, no mutable state, no general-purpose policy language) — are stated in capability-language-core-principles-v1.md (in ):

• P1 minimal core · P2 no control flow · P3 immutable values · P4 domains, not types · P5 deterministic and terminating · P6 fail-closed · P7 define once, consume everywhere · P8 carriers separate from semantics · P9 local decidability · P10 bounded work · P11 composition narrows only · P12 ≥2 independent implementations

Terminology Term Definition Action The thing being referenced — an abstract operation class (auth) or a concrete asserted effect (evidence). CapabilityId Structured name identifying a class of actions within a scheme. Operation Concrete action request: a CapabilityId plus parameters. Grant Principal's authorization of a CapabilityId with optional params and constraints. Binding Abstract concept connecting an Identity to an Action. Entailment (auth) and Match (evidence) are concrete instances. Entailment Authorization-side binding: "Grant G covers operation O" (⊆). Match Evidence-side binding: "Evidence E is bound to exact action A". Constraint Bound on how an action may be used (auth) or what evidence is required (evidence). Intersection Combining multiple grant sources into an effective set (∩). Verdict Outcome of evaluation: allow/deny/allow_unresolved (auth) or SATISFIED/UNSATISFIED (evidence). Decision Authorization-side verdict: allow, allow_unresolved, or deny + reason (+ additive unresolved). Satisfaction Evidence-side verdict: SATISFIED or UNSATISFIED. Note: Binding in CLC-v1 denotes the identity↔action relation (coverage on the authorization side, match on the evidence side). It is not key binding (cnf / DPoP / mTLS sender constraint), which belongs to the native artifact's specification (see Section 11). Verdicts are written lowercase on the authorization side (allow/deny/ allow_unresolved) and uppercase on the evidence side (SATISFIED/UNSATISFIED), following the EMILIA/AEB convention.

Grammar

capability-id = scheme ":" action [ ":" wildcard ] wildcard = "*" scheme = vendor "/" product "-v" major vendor = 1*( ALPHA / DIGIT / "-" ) product = 1*( ALPHA / DIGIT / "-" ) major = 1*DIGIT action = segment *( ":" segment ) segment = 1*( ALPHA / DIGIT / "-" / "_" / "." )

The trailing wildcard is part of the identifier grammar (the optional final ":" "*" above), so std/database-v1:query:* is a well-formed capability-id; only that shape is a wildcard (Section 3, below). Unambiguous scheme. Because product may contain -, the -v<major> suffix is the last occurrence of -v followed by digits. A product that itself contains the two-character sequence -v is invalid (invalid_capability_id), so a/b-v1-v2 is rejected rather than parsed two ways; b-v1 as a product name is not expressible. A scheme in use before this rule that relies on such a product is out of grammar in v1. A v1 identifier is scheme:action (e.g. std/database-v1:query:SELECT). Trailing * as a complete segment matches one or more remaining segments (it does not match the empty remainder: std/database-v1:query:* does not cover std/database-v1:query). Wildcard scope for v1: CLC-v1 core defines exactly one wildcard shape: the complete trailing segment *. Partial (std/crm-v1:re*), bare *, **, {a,b}, [a-z] are reserved for v2; a v1-conforming implementation MUST reject them (unsupported_wildcard). If a capability scheme declares its own extended grammar, that scheme's conforming implementation may accept it; but CLC-v1 core conformance neither requires nor authorizes those forms (see Section 12). Wildcard detection precedes grammar conformance. A forbidden wildcard shape is reported as unsupported_wildcard even when the same string also violates the base grammar (Section 3 segment): the wildcard-shape checks run before the generic invalid_capability_id test. Examples: std/database-v1:query:SELECT valid database:query invalid std/database-v1:query:* valid std/database-v1:query:SEL* invalid

Action An action is the thing being referenced. CLC-v1 defines two concrete forms:

Operation (authorization side) A concrete action request: CapabilityId + parameters.

{ "id": "std/database-v1:query:SELECT", "params": {"tables":["customers"], "limit":{"max":50}} }

Missing id → deny("missing_capability_id").

ObservedAction (evidence side) The material action constructed by the effect boundary from executor-controlled facts. CLC-v1 defines the interface, not the construction algorithm. An ObservedAction carries: action_type: the action type name declared by the relying-party-pinned type definition (the class this projection belongs to) material_fields: every field the type definition declares material digest: computed over the canonical material projection (below) The material projection is deterministic and normative: The action type declares a material field set (required and optional-but-included). Only that set enters the digest. Canonical serialization is the JSON Canonicalization Scheme (JCS) ; the one suite defined in v1 is jcs-sha256 (Section 4.3). The projection identity is the language's own, not a CAID: clc-action:1:<type>:<suite>:<b64url>. A CAID covers the complete Action Object under its own suite registry and identifies the action object, not an occurrence; this projection covers only the declared material set. The two are related by a relying-party-pinned Action-Mapping Profile (Section 6.4), never by treating the strings as interchangeable. A field the type does not declare as material MUST be excluded from the digest and MUST NOT affect Match: an ObservedAction carrying undeclared fields is not invalidated, but those fields carry no action identity. A type-declared material field that is missing makes the ObservedAction non-matchable: coverage MUST NOT be inferred, defaulted, or repaired (UNSATISFIED, Section 10). This is the evidence-side mirror of key closure (Section 6.2): the absence of a governing field is fail-closed, never fail-open. The effect boundary MUST construct the ObservedAction from facts it controls. It MUST NOT copy a requester-supplied action digest without deriving or checking the corresponding fact.

Identity Two identity levels, corresponding to the two Action forms: Level Identity Scope Example Class CapabilityId Covers a class of actions std/database-v1:query:* Instance ActionId (projection digest) Identifies the material content of one action, not an occurrence clc-action:1:payment.release.1:jcs-sha256:... A CapabilityId covers a class; an ActionId identifies the material content of one action. It does not identify an occurrence: ActionId binds the declared material content, and correlation to a particular occurrence additionally requires an occurrence discriminator defined and checked by the consuming profile. CAID-05 Section 4.7 carries such a discriminator as the optional occurrence_id of an action object; when a profile uses one, it MUST appear among the declared material fields for it to affect the digest. Allocating unique occurrences and proving one-time consumption or execution stay outside both documents (CAID-05 Section 7). Entailment checks class coverage; Match checks content binding.

Grant A grant = CapabilityId + optional params + optional constraints.

grant = capability-id [ params ] [ constraints ] constraint = scheme ":" type [ ":" params ]

Constraints are deny-when-declared: explicitly empty bound (e.g. zero max, empty allowlist) denies the class; omitted bound uses scheme default.

Binding Binding is the abstract concept of connecting an Identity to an Action. CLC-v1 defines two concrete instances:

Entailment (authorization binding) Grant G covers operation O if: Literal: G = O (byte-for-byte after normalization). Trailing wildcard: G = scheme:prefix:* and O has at least one segment after scheme:prefix: (segment-boundary comparison, not lexical prefix). Grant Operation Result std/database-v1:query:* std/database-v1:query:SELECT yes: wildcard matches std/database-v1:query:* std/database-v1:query:SELECT:deep yes: matches multi-segment std/database-v1:query:* std/database-v1:admin:DDL no: different namespace std/database-v1:query:SELECT std/database-v1:query:INSERT no: literal mismatch

Parameters Type Rule Example number op ≤ grant 50 ≤ 100 (holds) string exact "a" = "a" (holds) boolean exact true = true (holds) array set of allowed values (enum): request scalar must equal a member; request array — every element must equal a member {"station":[1,2,3]} ⊇ 2 (holds); ⊇ 9 (does not hold) object every grant key in op, values recurse {"t":["id"]} ⊆ {"t":["id","name"]} (holds) other exact equality — Boolean parameters are exact only. A boolean grant value is matched by exact equality (true covers true only). Booleans are not numbers: an implementation MUST NOT interpret true/false as 1/0 and MUST NOT run them through the numeric-bound rule (op ≤ grant). They also play no role in numeric comparisons or bounds. (Boolean GRANT-side bounds are rare in practice; the setting that matters — an operation-side boolean value under key closure — is pinned by undeclared-001.) Table rows map to vectors: number → params-001/params-002; array scalar member → params-009 (holds) / params-010 (fails); array element-wise → params-011 (holds) / params-012 (fails); object recursion allow → params-015, deny-direction → params-005; categorical guard → params-014. Under the old (pre-v1.1) array-as-bound rule the object example read (does not hold); with the v1.1 enum rule the request-side element id is a member of the granted set, so the row is (holds) (see clc-v1-ambiguities.md Section 2). Enum semantics of arrays (v1.1). An array-valued grant parameter is the set of allowed values, not an order. The request MAY supply a single scalar (must be a member of the set) or an array (every element must be a member). Members compare by exact equality: a number inside a granted array denotes that exact value, not a bound. This is what keeps categorical identifiers safe — grant {"station":[3]} does not cover station 2 (failure → deny("not_in_enum")). Scheme authors SHOULD encode categorical parameters (station, cell, batch, tool id) as arrays; a scalar number in the grant keeps upper-bound (bound) semantics for ordered quantities (max_rows, speed, angle). An explicitly empty grant array [] denies the class (empty_bound_denies_class). Grant with no params — or with an empty {} params object — covers any operation params (unconstrained). An absent params and "params":{} are semantically equivalent — consistent across entailment (Section 6.3) and intersection (Section 7 rule 6; Section 9.1). A null parameter value is invalid in v1 → reject (invalid_params_null). Absent ≠ explicitly empty (see Section 8). Key closure (Plan A). A bounded grant governs its declared keys: an operation carrying a parameter key the grant does not declare is rejected → deny("undeclared_param") (fail-closed, Section 9.1 layer 7 request side). An unconstrained grant (no params, or params:{}) accepts any operation params, so no key closure applies there. Within layer 7 the missing-key check (params_missing) resolves before the undeclared-key check (undeclared_param); both are key-level and run before the enum/bound value checks (layers 8–9). Params representation and input normalization (v1.1). Before any Section 9.1 layer runs, the params object is normalized at the input boundary: Canonical serialization. Params are normalized to the JSON Canonicalization Scheme form (JCS, RFC 8785): object members sorted by code unit, numbers in the ECMAScript Number::toString form, no insignificant whitespace. The original key order, number spelling and spacing are not preserved — canonicalization replaces them with the deterministic form (this is what the size cap in step 4 and the material digest are measured on). Evaluation never guesses; a lossy re-serialization (e.g. a map that drops duplicate keys) is not used for decisions. Duplicate keys. A params object with a duplicate JSON key is rejected: deny("invalid_params_duplicate_key"). Number shape. A numeric param is rejected with deny("invalid_params_number") when it is non-finite or out of IEEE-754 range (e.g. 1e400, an exponent beyond the representable maximum — it is an overflow, not an over-precision case) or over-precision (more than 17 significant decimal digits, e.g. 1.0000000000000001). The digit count operates on the JSON token as received — the raw digit-character sequence at the input boundary, before it enters any storage / float / decimal representation — so float64 and decimal/bignum implementations MUST NOT diverge: the judged input is always the raw token text (params-* number probes in the corpus; near-limit forms such as 1.0000000000000001 extend the probes without changing the rule). Number identity is the JSON number, not its spelling. Two tokens that denote the same JSON number are one value to the layers: -0 and 0 (JCS renders both as 0), 1e3 and 1000, 1.5e-3 and 0.0015, 1.0 and 1e0. The layers compare the canonicalized number, never the received token text. Integer bound (I-JSON). An integer-valued number whose magnitude exceeds 2^53 − 1 (9007199254740991) is refused with the same code — fail-closed, never rounded. 2^53 − 1 is the top of the I-JSON interoperability range ( Section 2.2): it is the largest integer that both has an exact IEEE-754 binary64 representation and is guaranteed to round-trip, whereas a larger integer-valued number may be silently rounded to a neighbour (9007199254740993 → 9007199254740992). The bound is judged on the mathematical value of the received token, independent of its spelling and of whether the port's carrier could hold it exactly — so 9007199254740992 (2^53) is refused although binary64 represents it exactly, and 1e100 is refused although binary64 cannot represent it exactly: an acceptance rule that turned on binary64 representability would let the numeric carrier a port happens to use change the verdict. It is not judged on digit count (9007199254740993 has 16 digits, so the over-precision rule above does not catch it). param_bounds numerics inherit the bound through Section 6.5's input-normalization clause. Non-integer numbers. A non-integer-valued token of at most 17 significant digits is not reached by the integer bound; it arrives as binary64 through step 1 (IEEE-754 round-to-nearest, ties to even), whose result is deterministic across implementations. A port whose carrier holds more precision than binary64 (a decimal or arbitrary-precision type) MUST compare the value by that same binary64 value and MUST NOT retain precision binary64 cannot hold — otherwise the carrier, not the language, would decide. Malformed params text that cannot be parsed as an object is refused with the same code (see step 7). Size and depth. Params whose canonical length exceeds 512 UTF-8 octets — measured in octets, never in code points or UTF-16 code units (rev CLC-1.4) — or whose nesting depth exceeds 32, are rejected: deny("invalid_params_size"). Nesting depth counts both objects and arrays, with the outermost object as level 1 (31 nested arrays inside the top-level object therefore = depth 32). Order of checks. Size/depth (4) precede duplicate keys (2), which precedes number shape (3); the first failing check wins. All five run before Section 9.1 layer 1, so normalized params are the only view the layers see. The size in step 4 is measured with every token in its JCS spelling (numbers re-spelled per ECMAScript Number::toString, strings per JCS Section 3.2.2.2) but without requiring key uniqueness, so it is defined even when a duplicate key makes full JCS undefined; this is why an over-limit input that also has a duplicate key reports invalid_params_size (params-025/params-026). Decoded-object path. A caller that supplies params already decoded (no raw_params text) cannot reproduce the original byte stream; in that case the size check (4) applies to a canonical serialization (the JCS form of the decoded object: sorted keys, compact), and the depth check (4) applies to the decoded structure directly. Byte-exactness against a specific original text is guaranteed only for the raw path; but both entry points MUST reject the caps — an oversized/deep params object is denied whichever way it arrives. Malformed Unicode is a property of the received text. A lone surrogate escape, or an octet that is not valid UTF-8, has no JCS form (Section 3.2.2.2): step 1 cannot normalize it, so the input is refused. The refusal reuses invalid_params_number (Section 9.2), whose scope is "the params input is not representable in canonical form" — a numeric param that is non-finite / over-precision or text that is not well-formed Unicode / not parseable as a JSON object. (One stable code, not one per malformation: introducing a second code would change the reason for inputs already pinned to invalid_params_number, which the additive rule forbids.) That check is defined over the octets as received and MUST run before any decoding step, because a general-purpose JSON decoder does not preserve the distinction. Go's encoding/json, for example, replaces both a lone surrogate escape and an invalid octet with U+FFFD, so {"s":"\ud800"}, {"s":"\ufffd"} and a literal invalid octet arrive at the decision function as the same value — {"s":"\ufffd"} — and the refusal cannot be recovered afterwards. Therefore: the party that receives the input MUST run steps 1-5 on the received text, not on a re-serialization of a decoded value; the entry point that takes the text is the sole conformance entry point: conformance (Section 12) is judged there, and an implementation SHOULD name the two entry points so a caller cannot mistake one for the other; a decoded-value entry point inherits the obligation of the received text: the decoding that produced the value MUST have been lossless — a decoder that repairs malformed input (replacing an invalid octet or an unpaired surrogate escape with U+FFFD, as Go's encoding/json does) MUST NOT feed an entry point used for a conformance claim, because the repaired value can no longer reproduce the refusal the received text requires. An implementation whose decoded entry cannot establish that its decoding was lossless MUST NOT be described as refusing malformed Unicode: its verdict is defined over the value it was handed, which may already be a repaired one; two parties that must agree on the verdict MUST agree on the received text, or on a digest of it. Byte-order mark is refused (raw path). A leading UTF-8 BOM (EF BB BF, U+FEFF) is not part of a JSON text: Section 8.1 permits a parser to ignore it, but this specification MUST NOT — params text that begins with a BOM is refused with deny("invalid_params_number") before parsing. The refusal outranks the step-4 caps: a BOM-prefixed text that is also over the size/depth limit reports invalid_params_number, never invalid_params_size (params-062 pins this precedence). An implementation MUST NOT strip U+FEFF as whitespace: ECMAScript classifies it as WhiteSpace (so a JavaScript trim() pre-step silently accepts a BOM-prefixed text that Python's str.strip and Go's strings.TrimSpace both refuse) — this rule closes that divergence fail-closed. A BOM inside the text is likewise invalid (U+FEFF is not a JSON structural character). Host numeric carriers normalize to JSON number semantics. A numeric value that reaches a comparison through a host-language numeric type — an integer type vs. a float type (int vs. float64), an arbitrary-precision or decimal carrier (BigInteger, BigDecimal, decimal) — MUST be compared under the JSON number semantics of steps 1–3: the same JSON number yields the same verdict whichever host type carries it. A numeric carrier MUST NOT fall out of the numeric branch into a string-shaped fallback comparison, and a host equality MUST NOT be applied that coerces across JSON types (Section 6.5 layer 8: true ≠ 1, "1" ≠ 1). Where a host offers a carrier binary64 cannot hold, the step 3 integer bound decides acceptance before any carrier-specific arithmetic; a value that cannot be compared losslessly fails closed (invalid_params_number) rather than being silently rounded. The property is not observable from the JSON corpus alone (every JSON number decodes to one host type per implementation); it is checked by representation invariance across the port's own numeric types, and params-040/params-041 guard the contract at the language level.

Algorithm

Entails(G, O) → bool: 1. G.namespace ≠ O.namespace → false (namespace = scheme + action Class, Section 9.1 layer 3) 2. G.id doesn't cover O.id → false (path coverage, Section 9.1 layer 4) 3. G declares no key → true (both `params` and `param_bounds` absent/empty → grant unconstrained, absent ≡ empty object) 4. O.params absent → false (bounded grant, request omits it → fail-closed, except a declared key marked `optional`, Section 6.5) 5. params_subset(O.params, keys(params) ∪ keys(param_bounds), param_bounds) (compare declared set incl. Section 6.5 bounds, Section 9.1 layers 5–9)

Step 5 is the Section 9.1 layers 5–9 comparison over the grant's declared key set (keys(params) ∪ keys(param_bounds), Section 6.5): params values follow Section 6.2 value subset, and a param_bounds key additionally enforces its Bound (inclusive min/max, step, enum cardinality, optional, nested). A grant with no param_bounds reduces step 5 to the params-only comparison of earlier revisions. Steps 3 and 4 read the declared key set, so a grant whose only declarations are param_bounds is still "bounded" for presence purposes. When more than one check fails, the reported reason follows the fixed resolved-reason ordering of Section 9.1. The remaining rules of this section are unchanged. Rule for missing operation params: If the grant has a params bound (step 3 does not apply) and the operation has no params field at all (not merely a key missing, but the entire field absent), step 4 applies: Entails → false → deny("params_missing"). This covers both "operation omits the entire params object" and "operation omits a single bounded key" — both fail closed. If a capability scheme declares a default value for a parameter, the implementation MUST apply that scheme default to O before step 4; absent such a declared default, step 4 denies. Layer 6 (null) resolves before presence. If the grant's params carry a null value (or the operation's params do), the failure is invalid_params_null (Section 9.1 layer 6) and is reported even when the operation omits the params field entirely — i.e. layer 6 resolves before step 4's params_missing. This is the fixed Section 9.1 ordering; it shadows the literal step order of the algorithm above, which lists presence before the null checks.

Match (evidence binding) Evidence E is bound to exact action A if: E carries a valid ActionId in the language's projection form (clc-action:1:…, Section 4.3). A CAID is a different object and relates to it only through a pinned Action-Mapping Profile. E's ActionId equals the recomputed ActionId of the ObservedAction. The ActionId was computed under the relying-party-pinned suite and definition source. Match is content correlation only. It does not validate a native artifact and does not authorize execution. It also does not identify an occurrence: correlation to a particular occurrence additionally requires an occurrence discriminator defined and checked by the consuming profile (CAID-05 Section 4.7), and a profile that uses one MUST pin how it is obtained and that the language sees it among the declared material fields. Cross-format mapping (E's native format ≠ A's canonical form) uses an Action-Mapping Profile: a hash-identified projection pinned by the relying party, with results EQUIVALENT_UNDER_PROFILE, NOT_EQUIVALENT, or INDETERMINATE.

Extended parameter bounds (param_bounds) params (Section 6.2) stays exactly as defined: a scalar number is an upper bound, an array is a membership set, an object recurses per key, and there is no lower bound, no step, no cardinality bound and no optional-key marker. This subsection adds those as a separate, optional grant field, param_bounds, so that no existing params input changes meaning: a grant without param_bounds behaves exactly as before, and every params verdict is unchanged. Binding rule (one authoritative representation per key). A parameter key MUST be declared in at most one of params and param_bounds. Declaring the same key in both is rejected: deny("invalid_params_binding"). The declared key set of a grant is keys(params) ∪ keys(param_bounds); key closure (Section 9.1 layer 7) and the params:{}≡absent rule are read over that set. Four distinct layers — do not collapse them in a decoder. A bare {} has different meaning at each layer, and an implementation must keep them separate before normalizing: Container presence — "params" absent vs "params":{}. Both are unconstrained (no declared keys); {} is not "declares nothing and denies". Declaration site — a key is declared via params or param_bounds (never both). The site decides which value algebra applies to the key. Value constraint — inside params, an empty value is a restriction ({"tables":[]} denies the class; deny-when-declared, Section 6.2); inside param_bounds, an empty Bound {} declares the key with no value constraint (it only participates in key closure). Same {}, opposite reading, because the layer differs. Request value — what O supplies (or a materialized default), evaluated per the key's family at layers 7–9. A decoder that maps {}→"absent" or {}→"empty set" uniformly across these layers will disagree with the language on at least one of them. Bound grammar (closed). param_bounds maps a key to a Bound object:

Bound = { // a closed JSON object, at most one family // numeric family "min": <number>, // inclusive lower bound "max": <number>, // inclusive upper bound "step": <number > 0>, // value must be an integer multiple of step // enum family "enum": [ <scalar> ], // allowed values (membership set) "min_items": <integer >= 0>, // request cardinality lower bound "max_items": <integer >= 0>, // request cardinality upper bound // nested family "nested": { <key>: Bound }, // recursion for an object-valued key // orthogonal to all families "optional": <boolean> // request MAY omit the key (default false) }

(Bound is JSON, not ASN.1: the members above are JSON keys and the comments are notation only.) Every member is optional and the object is closed — a member outside this set is rejected (invalid_params_binding). A Bound carries at most one value family, plus optional which is orthogonal: numeric family — any of min, max, step; enum family — enum, and/or min_items/max_items; nested family — nested. Mixing families (e.g. enum with max, or nested with min) is rejected (invalid_params_binding). An empty Bound {} declares the key with no value constraint (it still participates in key closure). min > max, step ≤ 0, min_items > max_items, or a negative min_items are rejected (invalid_params_binding). The whole param_bounds object is subject to the same input normalization as params (Section 6.2): canonical serialization, duplicate keys, number shape, size (512 octets) and depth (32), checked at layer 2. Entailment semantics (grant G vs operation O). Presence (layer 7). A declared key with optional: true MAY be absent from O; a declared key with optional absent or false MUST be present (params_missing otherwise). An operation key not in the declared key set is undeclared_param. (Unchanged behavior for grants that declare no param_bounds, where every declared key is required.) Enum family (layer 8). If enum is declared, the request value must be a member (not_in_enum, unchanged rule): a scalar request must equal a member; an array request must have every element equal to a member. Equal here is JSON type-sensitive equality: two values are equal only when they have the same JSON type and the same value — true equals neither 1 nor 0 (the Section 6.2 rule that booleans are never numbers applies to set membership as well), and the string "1" equals neither the number 1 nor true. Numbers are compared after the Section 6.2 canonicalization of the input boundary (one IEEE-754 binary64 value, rendered per ECMAScript Number::toString under JCS ), so 1 and 1.0 — the same value in two spellings — are equal. The same equality is used by Section 6.2 array membership, by the Section 6.6 enum intersection, and by Section 13.4.3 enum narrowing; implementations MUST NOT substitute a host-language equality that coerces across JSON types (e.g. a language where true == 1). The host's numeric carrier MUST likewise not decide the comparison branch (Section 6.2 step 9): every numeric carrier — integer, float, decimal, arbitrary-precision — compares under JSON number semantics. If min_items / max_items are declared, the request cardinality (an array's length; a scalar counts as 1) must satisfy min_items ≤ n ≤ max_items, else params_cardinality. Numeric family (layer 9). min/max are inclusive: a numeric request value must satisfy min ≤ v ≤ max (each bound, when declared), else params_out_of_range. step requires the request value to be an integer multiple of step, evaluated in IEEE-754 binary64 as let q = v / step in q == floor(q) && q * step == v (deterministic across implementations, since both operands arrive at the boundary as binary64), else params_not_multiple. A numeric-family bound applied to a non-number request value is fail-closed (params_exceed_grant). Nested family (layer 8/9). nested recurses the value comparison on an object-valued request key with the same rules, including symmetric key closure and optional at each depth. A nested bound applied to a non-object request value is fail-closed (params_exceed_grant). Scheme defaults. A capability scheme (Section 3 scheme grammar; data/std/...) MAY declare param_defaults, a map from a parameter key to its default value. The consumer obligation of Section 6.3 is thereby made concrete: before evaluating Entails(G, O), an implementation MUST materialize, for every key O omits that the scheme declares a default for, that default into O's params; precedence is explicit operation value > scheme default > absent. A default is applied only when it is needed to satisfy a grant-declared key; it never adds a key the grant does not declare (the key-closure check is unchanged). optional closes the default interaction (no recursion). The presence of the key is decided first, by the grant's optional marker, and only then is a default consulted — so there is no "is the default needed?" loop: a required declared key (no optional/optional:false) that O omits: materialize param_defaults[key] if the scheme declares one (the key is then present with that value), else deny("params_missing"); an optional:true declared key that O omits: the default is not materialized — optional is the explicit "may be absent" marker and wins over the default, so the key is simply absent (and satisfies presence); either way, an explicitly supplied O value wins over the default. Defaults are scheme configuration, not language semantics: the relation itself takes no scheme argument, and two consumers that load the same scheme reach the same verdict (a deployment that omits the scheme's defaults is a different input, and the difference is attributable to the input, not to Entails). In the containment relation a default is resolved before the grants are compared (Section 13.8.1 obligation 1), so containment compares resolved declared sets. A param_defaults entry is not part of a grant's declared set and does not enter the containment lattice: Contains narrows over keys(params) ∪ keys(param_bounds) and the optional markers alone, so a parent's default never widens or inverts the optional→required narrowing direction (Section 13.4.3). Containment (Section 13). Section 13.4.3 narrows a child grant against its parent using this grammar: a param_bounds bound of the child must be within the parent's bound for the same key (numeric min raised or equal, max lowered or equal, step refined to an integer multiple, enum a subset, cardinality bounds tightened, optional not widened from required to optional, nested recursed). A parameter declared in one of params/param_bounds and not the other is a key-set difference and fails key closure (params_not_narrower).

Intersection of bounds (BoundMeet) Intersect (Section 7) combines the param_bounds of its sources with a meet, over the same "value → authorization set" denotation Section 7 uses for params. The families are those of Section 6.5; the meet is defined per key declared in param_bounds by two or more sources. optional is orthogonal and combines by conjunction: the result key is optional only if every source marks it optional (optional := AND); a key required by any source stays required. numeric family (min/max/step): min is the greatest declared minimum, max the least declared maximum (undeclared means unbounded). step: if both declare one and one is an exact multiple of the other (the Section 6.5 multiple predicate), the meet's step is the coarser (larger) of the two — the coarser grid is a subset of the finer, so it is the meet. If neither declared step is an exact integer multiple of the other, the meet fails closed (invalid_params_binding). The two grids may still share values — the common grid of step:5 and step:7 is the set of multiples of 35, so the intersection is not empty — but the meet's step must be one of the two declared steps: neither 5 nor 7 is an integer multiple of the other, so neither declared grid is a subset of the other, and the language does not synthesize an undeclared grid (such as the least common multiple) — no single step member of the two Bounds denotes both grids. One declared step is carried through. min > max after combining → the meet is empty → no_overlap. enum family (enum/min_items/max_items): the enum member sets intersect (exact equality — the JSON type-sensitive equality Section 6.5 layer 8 defines, so 1, true and "1" are three distinct members); if both sources declare enum and the intersection is empty → no_overlap; min_items is the greatest declared value, max_items the least; max_items < min_items → no_overlap. nested family (nested): the two objects MUST have the same key set, else no_overlap (exactly as object-valued params, Section 7); the result recurses BoundMeet per key, with optional combining as above at every depth. numeric ∩ enum (either source order): refused — fails closed (invalid_params_binding). A sound meet would have to carry both the numeric side's shape constraint (min/max/step, which Section 6.5 layer 9 applies to numeric scalar request values only) and the enum side's member set (which Section 6.5 layer 8 applies to scalars and to arrays whose elements are all members) in one Bound — two families, which the closed Section 6.5 grammar deliberately does not express. Filtering the member list by the numeric bound (the rev CLC-1.14 rule, removed here) was broader than either source: for numeric {min:2,max:4} ∩ enum {enum:[1,3,5]} the filtered result enum{3} accepts the array [3] (Section 6.5 layer 8: every element is a member), while the numeric source fail-closes that same request (Section 6.5 layer 9: a numeric-family bound applied to a non-number request value is params_exceed_grant). Refusing the whole combination agrees with the rest of the language: Section 6.5 rejects declaring two families in one Bound ("Mixing families … is rejected (invalid_params_binding)"), and Section 13.4.3 rejects narrowing into a family the other side does not declare ("a child Bound that adds a family the parent does not declare … is params_not_narrower"). The refusal covers every numeric × enum pair — with or without a member list on the enum side (a cardinality-only enum is the same cross-family clash), and regardless of whether any member happens to fall inside the numeric range: the family clash is decided before any member or range math, and it is symmetric in the two sources. scalar ∩ nested (numeric or enum vs. nested), in either order: a cross-family meet, refused with invalid_params_binding before any value math — no single-family Bound can carry both a scalar shape (a value presence) and the object recursion; the refusal is symmetric in the sources and holds regardless of the nested side's contents, exactly like the numeric × enum case (design-notes D12). A CLC-1.14 draft read this pair as an empty meet (no_overlap, since no request value is both a scalar and an object); rev CLC-1.15 adjudicates it to the unrepresentable-meet code so that the empty-meet code stays reserved for genuinely empty meets within one value family. empty Bound {} is the identity for the value families ({} ∩ X = X); its optional still participates. The result always carries at most one value family, so it is a valid Section 6.5 Bound. All failure modes are the Section 9.2 codes already associated with Intersect (no_overlap for an empty meet within one value family, such as min > max or disjoint enums; invalid_params_binding for an unrepresentable one — every cross-family pair in either order, including scalar ∩ nested, and incommensurable step grids); no new reason code is introduced. Declared domains meet here too. Where a scheme declaration (Section 6.7) names a domain for a key, two values of the same declared domain meet per that domain's rule in Section 6.7 (quantity: the narrower normalized bound, same unit, unit_mismatch otherwise); a pair of different declared domains is a cross-family pair and refuses with invalid_params_binding, exactly as the scalar × enum and scalar × nested pairs above. Key site. A key's declaration site (params vs param_bounds) must agree across the sources of one Intersect. A key declared in params by one source and in param_bounds by another is refused (invalid_params_binding). This can only arise for independent authority sources, never for a delegation chain: Section 13.4.3 already requires the sites to match at every hop (a site difference is a key-set difference). A key that no source declares in param_bounds stays in params and follows the Section 7 params meet unchanged.

Scheme Declaration (normative) A scheme declaration states, for one scheme, the domain of each parameter. It is an input to evaluation, pinned by the carrier or the relying party; the core never fetches it (P9). The artifact's grammar is CLC-SCHEME-v1 (schema: scheme-declaration.schema.json); this subsection fixes what CLC does with it. Why it exists. Without a declaration the domain is inferred from the value's shape — number → bound, array → enum, scalar → exact (P4). That inference cannot express a quantity, a decimal string compared in its value space, a subdomain, or a path prefix. A declaration states the domain instead of leaving it to the shape. It adds no grammar: params and param_bounds are unchanged, and a parameter it does not name keeps the inferred domain, so every existing grant keeps its meaning. Domains. In addition to bound, enum and exact, this revision defines four domains. Each is a type, not syntax (Section 6 extension rule), so none of them adds a wildcard shape or an operator: Domain Value Comparison Containment (Section 13.4.3) quantity {"value": <number>, "unit": <string>, "scale"?: <integer>} the unit MUST equal the declaration's unit (no conversion, unit_mismatch otherwise); value × 10^-scale ≤ the granted bound same unit; child ≤ parent decimal-string a decimal literal matching ^[+-]?[0-9]+(\.[0-9]+)?([eE][+-]?[0-9]+)?$ — no surrounding whitespace, no underscores, no NaN/Infinity, no .5 or 500. value-space comparison ("500" ≡ "500.00", "1e3" ≡ "1000", "007" ≡ "7") child ≤ parent subdomain a DNS name — no surrounding or embedded whitespace, one optional trailing dot whole-label suffix, case-insensitive child is the parent or below it path-prefix a path segment-boundary prefix (/a/b does not cover /a/bc) child is the parent or below it Resolution order. Declaration validation (member closure, digest match, known domains) resolves before Section 9.1 layer 1 together with the Section 12.1 revision check. A malformed declaration, a digest mismatch, or a domain the implementation does not support is refused with unsupported_domain; a declaration that requires a later language revision is refused through the minor gate. An implementation MAY validate a declaration when it loads it, and the reference engine does; the evaluation path still refuses an unsupported domain on its own, so a caller that supplies a declaration through an API without that pre-validation cannot bypass the rule. Evaluation rules. Authorization uses the declared domain's comparison (Section 6.3); a value whose shape does not match the domain is invalid_params_binding. A parameter with a declared domain MUST be declared in params, not in param_bounds: the domain has its own comparison rule, and a param_bounds family on the same key would be a second, unspecified one. A param_bounds entry on a declared-domain key is invalid_params_binding. Contains compares child against parent under the parent's declared domain. If the two sides declare different domains for the same key, the pair is not comparable and the result is params_not_narrower. BoundMeet (Section 6.6) meets two values of the same declared domain per that domain's rule (for quantity, the narrower bound with the same unit; a different unit is unit_mismatch), and refuses a cross-domain pair with invalid_params_binding, exactly as CLC-1.15 does for cross-family meets. A declaration MUST NOT name a parameter the scheme's capability does not define, and MUST NOT name a domain outside the set this revision defines.

Intersection (∩) P_effective = P_principal ∩ C_agent ∩ P_gateway. Rules: Each source provides a grant set. Effective grant MUST be covered by at least one grant from every source. Same-capability constraints merged by union: every source's constraint strings are kept (constraints are conjunctive — each is independently checked, so a tighter bound of the same type binds by construction). The merge is a set union of normalized constraint strings, not a meet that discards a looser one; dropping a source's constraint would drop its restriction. Any source missing a capability → capability absent from effective set. Zero or absent sources fail closed. An intersection over no sources at all — an empty source list, or every source absent — has no effective set: deny("absent_source"). (A source that exists but carries no grant for the capability is rule 4, not rule 5.) Empty params declares no constraint. A source whose grant has a present-but-empty params object contributes no restriction. The accumulated bound is preserved: bounded-then-empty and empty-then-bounded must give the same result — source order MUST NOT change the outcome (consistent with direct-grant params:{} ≡ absent, Section 6.3 step 3). Identifier comparison is params-free (rule 2). The effective identifier is the narrowest one covered by every source, chosen by the Section 6.1 identifier rules alone. It MUST NOT be resolved by routing grants through Entails with params — a bounded grant faced with a params-less sibling would trigger presence handling (Section 6.3 step 4) and wrongly fail this open; the identifier is compared, params merge separately (rule 3). Property: composition narrows only. If Intersect succeeds, the result MUST be covered by every source and MUST NOT depend on the order of the sources; otherwise it MUST deny with a normative reason code and MUST never raise. This meet-law is what the Section 6.2 param-subset algebra preserves across sources; it is pinned by property-cases.json (1184 cases, Section 12). Notation in this section: params are JSON objects (e.g. {"tables":["a"]}); constraints use colon-notation triples (e.g. varwof/constraint-v1:time:window:[{"start":"00:00","end":"01:00"}]). The example tables below use compact shorthand for readability; each row shows the relevant params or constraints only. Deny-when-declared: {"tables":[]} (explicitly empty) = deny the class. Omitted = scheme default. A null value is invalid in v1 → reject (invalid_params_null). Canonicalization MUST NOT broaden (segment-boundary comparison, not lexical prefix). Source A Source B Result {"tables":["a","b"]} {"tables":["a"]} {"tables":["a"]} (holds) {"tables":["a"]} {"tables":[]} deny (unconstrained) (no grant) deny {"limit":100} {"limit":50} {"limit":50} (holds) (unconstrained) constraint time:window:[{"start":"00:00","end":"01:00"}] recognized, not evaluated by core (→ allow_unresolved + unresolved) {"tables":["a"]} {"tables":["b"]} deny The meet is over authorization sets, not JSON values. Neither Intersect nor Section 6.2 compares JSON values directly; each value denotes an authorization set and the meet is set intersection on that denotation. A number denotes (-∞, v] (an upper bound), an array denotes a membership set, a string/boolean denotes the singleton {v}, and an object denotes the product of its keys' denotations. So {"a":1} is not the set {1} but (-∞,1], and {"a":1} ∩ {"a":2} = (-∞,1] = {"a":1} — a minimum, not an empty set. A meet is empty (→ deny("no_overlap")) only when the denotations are genuinely disjoint: two enum sets with no common member (a numeric params meet is always a minimum, since params has no lower bound), or divergent key sets (next paragraph). Object-value intersection requires identical key sets. Two object values intersect key-by-key only when their key sets are the same; object values with different key sets → no_overlap deny. Merging "shared keys" would drop the keys the other source constrains, so the result would not be covered by that source (P11 composition narrows only): {"a":1} ∩ {"b":1} → deny(no_overlap). When the key sets are the same the values recurse under the Section 6.2 rules, so a numeric leaf intersects to its minimum ({"a":1} ∩ {"a":2} → {"a":1}, not a deny — matching the {"limit":100} ∩ {"limit":50} row above), and {"a":1} ∩ {"a":1} → {"a":1}. param_bounds merges by a defined meet. Intersect combines each key declared in param_bounds with the meet of Section 6.6: numeric min/max/step, enum member intersection and cardinality, and nested recursion, with optional combined by conjunction. An empty meet is no_overlap; an unrepresentable one is invalid_params_binding — any cross-family pair in either order (numeric × enum, including a cardinality-only enum and regardless of whether any member falls inside the numeric range; and scalar × nested), or two step grids where neither declared step is an exact integer multiple of the other (Section 6.6). A key declared in params by one source and param_bounds by another is refused (invalid_params_binding, Section 6.6 "Key site"). The Section 7.1 ConstraintUnion projection is separate and unaffected.

ConstraintUnion (derived projection) A consumer that has established a delegation chain often needs the chain's whole constraint burden without computing an effective grant. That is exactly the constraint projection of chained Intersect (rule 3), exposed as a named function:

ConstraintUnion(chain) → string[] // chain = ordered Grant[]

It returns the normalized union of every constraint string carried by the grants in chain — duplicates folded, result deterministically ordered. "Lexically sorted" is pinned to UTF-8 byte order: the normalized strings are compared octet by octet over their UTF-8 encodings. For well-formed Unicode text this equals code-point order, and it is identical in every implementation whatever the host language's native string representation is — notably it is not UTF-16 code-unit order, which places supplementary-plane characters (encoded as surrogate pairs, first unit U+D800–U+DBFF) before characters in U+E000–U+FFFF, so an emoji would sort before a private-use-area character although its code point is higher. This collation sits after Section 6.2 canonicalization, not instead of it: JCS Section 3.2.3's UTF-16 code-unit order governs object member order inside a canonical serialization, while this section's UTF-8 byte order governs the emitted constraint-string list; the two artefacts keep their own pinned collations and neither re-orders the other. The union is the same normalization Intersect applies, so the two never disagree. The residual-obligation list of Section 8.4 (and the ordered Resolve output of Section 8.5) re-uses this same collation; the collation is defined here once, not restated there. ConstraintUnion is a projection, not a meet: it does not compare identifiers or parameters, does not read or validate constraint values, and does not check containment. It asserts nothing about authority; the chain's per-hop Contains check (Section 13) and the operation's Authorize/Resolve (Section 9, Section 8.5) remain separate steps. Because it is not an effective-set computation, it carries no membership semantics of its own — an empty chain fails closed with absent_source (the same refusal Intersect gives over zero sources, Section 7 rule 5), so a caller that lost the chain cannot mistake "nothing to union" for "no constraints". This function is why Contains stays pure (Section 13.4.4): containment answers "is the child's declared boundary inside the parent's?", and the union answers "what does the chain collectively require?". Folding the union into Contains would make the relation no longer a subset on the declared tuple and would let a null constraint check hide a broken boundary.

Constraint Constraints are shared between authorization and evidence sides.

Authorization-side constraints Limit how a capability may be used. Constraints use colon-notation triples: scheme:type[:params], where type is the second :-delimited segment (scheme : type, optionally followed by : and params): varwof/constraint-v1:max_rows:100 — type max_rows, param 100 varwof/constraint-v1:time:window:[{"start":"00:00","end":"01:00"}] — type time, param window: array of UTC window segments (a single window = a one-element array). A scalar form (time:window:3600, the sliding-duration/freshness concept) is not a legal time:window value in v1 → invalid_constraint. varwof/constraint-v1:network:cidr:["10.0.0.0/8"] — type network, param cidr: JSON array (≤ 32 elements, each a legal IPv4/IPv6 CIDR string) v1 core recognizes constraints by (scheme, type) pair, not by type name alone: the identity of a constraint is two things — the declaring scheme and the type. The core recognizes exactly these pairs:

varwof/constraint-v1 : max_rows varwof/constraint-v1 : time varwof/constraint-v1 : network

Any other (scheme, type) — including foo/database-v1:max_rows — is not core-recognized: it is rejected as unknown_constraint (fail-closed), never handed to a core evaluator. This kills cross-scheme semantic pollution: the type name alone never selects an evaluator, so a scheme that defines its own max_rows cannot have its semantics hijacked by Core's max_rows evaluator (§P7 define once, consume everywhere — scoped per declaring scheme). Scheme-defined constraint types belong to the v2 / profile layer (evaluated by the declaring scheme); core's non-recognition of them is fail-closed (unknown_constraint). A recognized constraint MUST also conform to that type's value grammar (table below); a recognized type with a non-conforming value is rejected as invalid_constraint — never silently skipped, never passed through. Intersection itself neither evaluates nor validates constraint values (it only merges constraint strings; the Section 8.1 value grammar is enforced at the decision boundary, Section 9, and in the constraint merge rules below). The core defines an evaluator for max_rows only; evaluation of time/network bounds is the declaring capability scheme's responsibility (the scheme evaluates, the core owns only what the Section 8.1 value grammar is, not whether a moment/address currently hits it; Section 11). A recognized-but-unevaluated constraint is carried on the decision's additive unresolved field — never silently dropped (Section 8.4). A non-recognized (scheme,type) → deny("unknown_constraint") (fail-closed). max_rows meets the Section 6.2 key closure. The operand of the max_rows constraint is the operation's max_rows parameter, read from the same params object layer-7 key closure governs. The parameter is not exempt from closure: an operation carrying max_rows against a grant that declares other params without declaring max_rows fails with undeclared_param: max_rows before the constraint is consulted, and an operation omitting the parameter under a max_rows constraint fails closed with max_rows:violated (table below; Appendix B D21). A grant that therefore declares any params and carries a max_rows constraint MUST declare max_rows among its params/param_bounds keys — the bound is written into both places at once (the constraint carries the enforced value, the declaration satisfies the closure), and the two checks apply conjunctively. Only an unconstrained grant (no params, or params:{}) can carry the constraint alone. A max_rows constraint value whose integer magnitude exceeds the Section 6.2 step 3 integer bound (2^53 − 1) is out of grammar → invalid_constraint (the same fail-closed bound; no silent binary64 rounding). type value grammar (v1) core behavior max_rows strict non-negative integer (JSON number grammar: no leading +, no 0x, no trailing characters) evaluated against op params; the operation-side value domain is a finite non-negative integer — an absent param, a string, a boolean, a negative, a fractional or a non-finite value, or a value above the bound → max_rows:violated (cannot be shown conforming = fail-closed; rev CLC-1.4); constraint value out of grammar → invalid_constraint time (window) non-empty JSON array (≤ 32 elements), elements {"start":"HH:MM[:SS]","end":"HH:MM[:SS]"}, UTC, repeated daily, single window = one-element array; each endpoint is a seconds-of-day value (parsed from HH:MM[:SS]), with the reserved endpoint "00:00" in the end position read as 86400 (next-day midnight, 24:00) — so a segment is the half-open [startSec, endSec) and MUST satisfy startSec < endSec (not lexical string order, which would wrongly reject the canonical 22:00→00:00 segment, since "22:00" > "00:00" as text); a single segment must not cross midnight, so a crossing window is split into two same-day segments (22:00→00:00 + 00:00→06:00); the segment list MUST be ascending by (startSec,endSec) and non-overlapping recognized only → allow_unresolved + unresolved (Section 8.4) network (cidr) legal IPv4/IPv6 CIDR strings (addr/mask), syntax-level check only; JSON array ≤ 32 elements recognized only → allow_unresolved + unresolved (Section 8.4) The expressible window set has no empty window and no full-day window; a declaring scheme that needs such windows extends the grammar (Section 11). Merge rules in v1: numeric → minimum wins; allowlist → intersection; unknown → deny. The allowlist is the array-enum set (Section 6.2); intersecting it takes the shared members. v1 core defines no denylist constraint — merging applies only to the varwof/constraint-v1 types the core itself evaluates; it does not define or merge scheme-defined types (Section 11). Constraint-set merging is always normalized and deterministically ordered (duplicate strings collapsed, result ordered) — the same input yields the same constraint sequence in any implementation. v1 does not tighten time/network bounds at intersection: intersection keeps only the constraint strings it encounters (no evaluation, no tightening, Section 8.4).

Evidence-side constraints Require specific evidence properties: freshness, consumption, quorum, initiator-exclusion, etc. These are structurally identical to authorization constraints — a type plus optional params — but evaluated against evidence artifacts rather than operation parameters.

Unified constraint grammar

constraint = scheme ":" type [ ":" params ]

Both sides use the same grammar. The validator (authorization) or evidence evaluator (evidence) interprets the type-specific params.

Recognized but not evaluated (residual-obligation channel) A recognized constraint whose value conforms to its Section 8.1 value grammar but for which the v1 core defines no evaluator — v1: time, network; evaluation belongs to the declaring scheme (Section 11) — MUST NOT be silently dropped: it must appear in the decision's additive unresolved list (Section 9), and the verdict must be allow_unresolved (not allow):

Decision = { verdict: "allow"|"deny"|"allow_unresolved", reason, unresolved: string[] }

Fail-closed boundary: allow_unresolved is an independent enum value, never equal to allow. A consumer (PEP / profile / declaring scheme) MUST evaluate or confirm every unresolved constraint before allowing; if it cannot execute or confirm, it MUST deny (AAC Section 6.6: "when it cannot perform or confirm, it should be treated as deny"). A consumer that only writes if decision.verdict == "allow" cannot, on the literal enum, release a residual obligation as an already-satisfied allow — a path that leaves obligations unconfirmed must explicitly handle allow_unresolved to pass. "the caller is supposed to check unresolved" is not sufficient defense: the verdict itself must refuse the two-value short-circuit. unresolved semantics and ordering: [] for deny and for fully evaluated allow; for allow_unresolved the normalized constraint strings with duplicates folded — the conditional collation order of Section 7.1 (#-separated, then by UTF-8 byte sequence), the same comparison the ConstraintUnion payload uses (rev CLC-1.15). The order is deterministic: the same input yields the same byte sequence in any implementation — in particular NOT the ECMAScript default string order (UTF-16 code-unit order); join the residual obligations across all sources and sort exactly once, before emitting the decision. Combined obligations (consumer side): multiple unresolved constraints of the same (scheme,type) form a conjunction (AND) — satisfying A and B satisfies all; OR, any-one, first-wins, and ignoring some entries are all forbidden. Constraints of different (scheme,type) do not interact; each evaluates under its own declaring scheme (P11 composition narrows only: conjunction only narrows). The core owns only the value grammar ("what it is"); "whether this window/CIDR currently forms a boundary" ("how to evaluate") belongs to the declaring scheme (Section 11) — but the boundary-moment semantics are part of the grammar (Section 8.1: half-open [start, end), single segment within one day, crossing split into segments), and a scheme MUST NOT change that interpretation, only evaluate on top of it.

Resolving residual obligations (Resolve) Section 8.4 delivers residual obligations but leaves the consumer's feedback loop undefined. Resolve closes it: a consumer reports, per obligation, whether it is satisfied, violated, or still unknown, and the core collapses the result to a fresh decision.

Resolution = { constraint: string, status: "satisfied" | "violated" | "unknown" } Resolve(decision, resolutions, now?) → Decision

decision is a Section 9 Decision; resolutions is an ordered list (possibly empty) of Resolution; now is an optional UTC instant (RFC3339, Z). now is the only input that lets the core tick a residual itself. Algorithm, in order: Terminal verdicts are fixed. If decision.verdict is deny or allow, Resolve returns decision unchanged and ignores resolutions: a deny is never revived and an allow carries nothing to discharge. Malformed input fails closed. An entry that violates the grammar above (unrecognized status, empty/non-string constraint) → deny with invalid_resolution; a now that is not a valid instant → deny with invalid_timestamp. Discharge on allow_unresolved. Let O = decision.unresolved (already normalized and sorted, Section 8.4). For each o ∈ O, a status in {satisfied, violated, unknown} is computed by combining the sources below under the order violated ≻ satisfied ≻ unknown (the most restrictive source wins — a consumer that reports violated is never overridden by a lenient one, and the core clock is never overridden by a lenient assertion): Core clock — if o is a core-recognized varwof/constraint-v1:time obligation whose value is a Section 8.1 window array, and now is supplied, the core evaluates it: now inside a segment → satisfied; outside every segment → violated. This is the obligation's TTL: the discharge horizon is the end of the segment containing now, so re-invoking Resolve with a later now re-evaluates to violated once the window has passed — a cached allow does not outlive its window. Consumer resolution — a Resolution for o contributes its status; the most restrictive of the two wins. Absent everywhere → unknown. Repeated entries for the same constraint combine under the same violated ≻ satisfied ≻ unknown order. Unrelated resolutions are ignored. A Resolution whose constraint is not in O cannot widen the outcome; O is authoritative. Result: any o has status violated → deny, reason = that constraint's {type}:violated (time:violated for a core-clock violation; otherwise the type segment of the constraint string), unresolved = []; every o is satisfied → allow, reason null, unresolved = []; otherwise → allow_unresolved, reason null, unresolved = the still- unknown subset (normalized + sorted, Section 8.4). Without now, the core attaches no TTL to a time:window obligation: a consumer may report it satisfied (its declaring scheme owns its clock), but that discharge is only as fresh as the call, and a consumer that needs a core-enforced horizon MUST pass now. Acting on a result (caching rule). Resolve returns a plain decision; it does not carry an expiry field, so a core-clock discharge is only valid for the instant of the call. A consumer that acts on a Resolve result involving a core-evaluated time:window obligation MUST either re-invoke Resolve with a current now immediately before acting, or not cache the result beyond the current segment's end (the discharge horizon, Section 8.5 step 3) — it MUST NOT hold a cached allow past that horizon. Equivalently: a suite that needs a mechanically enforceable "valid until" value derives it from the segment end itself, not from the returned Decision. (The obligation carried on an allow_unresolved decision remains the Section 8.4 object a consumer evaluates; this rule is only about not outliving a clock-based discharge.) Resolve is deterministic and fail-closed; it is idempotent (Resolve(Resolve(d, r, now), r, now) = Resolve(d, r, now)) and monotone (adding a violated never turns a deny into an allow; removing one never turns an allow into a deny). It neither invents nor drops obligations. It subsumes the coarser identity-level consumer gate (the reference implementations' Discharge, which only answers "do I understand and commit to every obligation?"): confirming an obligation is the satisfied case. It defines no policy about how the consumer evaluates an obligation (P2) — that is the declaring scheme's (Section 11).

Decision Function (Authorization Side)

Authorize(grants, operation) → Decision Decision = { verdict: "allow"|"deny"|"allow_unresolved", reason: string|null, unresolved: string[] } // additive, Section 8.4

Algorithm. One precedence rule governs the whole function: the grant-side pre-check resolves before operation validation. An absent or empty grant set denies with capability_not_authorized even when the operation is absent as well (Section 9.1 layer 10; Appendix B.5, row D15, pins it), so a caller that passes neither input gets that reason and not missing_capability_id. The algorithm below therefore runs the grant-side pre-check first, ahead of the numbered operation steps — the numbering mirrors Section 9.1's layer order for what follows, not the precedence of the pre-check. (Pre-check) An absent or empty grant set → deny("capability_not_authorized") (Section 9.1 layer 10), before any operation check; this is why the absent-grant / absent-operation pair resolves here and not to a layer-1 code. Validate operation: missing/invalid id → deny with stable reason. The operation's specific layer-1 code is reported — missing_capability_id (no id), unsupported_wildcard (v1-forbidden wildcard shape) or invalid_capability_id (other grammar violation) — and is not collapsed to a generic code. Find covering grants via Entailment (Section 6.1). No covering grant → deny("capability_not_authorized") (Section 9.1 layer 10); the absent/empty case was already resolved by step 0. For each covering grant: evaluate constraints (Section 8.1): non-recognized (scheme,type) → unknown_constraint; recognized but non-conforming value → invalid_constraint; recognized with a core evaluator (varwof/constraint-v1:max_rows) and violated → {type}:violated; recognized without a core evaluator (time, network) → residual obligation (Section 8.4). Aggregate (multi-grant set, Section 9.1): Any covering grant that "allows" (no params/constraint rejection) → overall allow; Residual obligations = the unresolved union across the covering grants that also allow (normalized + sorted). A covering grant that is rejected at the params/constraint layer contributes no obligations: it does not authorize, so its residuals are not carried. The union is nevertheless order-independent, because every covering-and-allowing grant is visited; When no covering grant allows: if at least one covering grant rejects at the params/constraint layer → use the rejection reason of the first covering grant in canonical order (deterministic, Section 9.1); if no grant covers at all → capability_not_authorized. Allow with non-empty residual obligations → verdict = allow_unresolved; allow with empty obligations → verdict = allow. The grant set may be absent or empty — e.g. an integrator calls Authorize with no grant value or with an empty capability record. A safe evaluator MUST NOT raise; it MUST resolve such input to deny("capability_not_authorized") (falling out of layer 10/step 2). An absent operation resolves to deny("missing_capability_id") (layer 1); the evaluator MUST NOT raise there either. Properties: deterministic (same input → same output), fail-closed, stable reason codes.

Resolved Reason Ordering (normative) When more than one condition fails for a grant/operation pair, the resolved reason (the single code reported) is the first applicable layer in this fixed order. It applies to Entails (Section 6.3), Intersect (Section 7) and Authorize (this section). Declaration validation resolves before layer 1. The language-revision check (Section 12.1) and the scheme declaration (Section 6.7) are evaluated before any layer below: an input whose declaration is malformed, whose digest does not match, or which names a domain this implementation does not support is refused with unsupported_domain (or unsupported_language_revision where the revision gate applies), and no layer-1..11 code is reported. # Layer Checks Reason code(s) 1 CapabilityId validity Section 3 grammar, wildcard shape invalid_capability_id, missing_capability_id, unsupported_wildcard 2 Params normalization Duplicate JSON keys; non-normalizable numbers (non-finite / out-of-range / over-precision) and malformed-Unicode text; size/depth limits (Section 6.2); param_bounds well-formedness and the one-representation binding rule (Section 6.5) invalid_params_duplicate_key, invalid_params_number, invalid_params_size, invalid_params_binding 3 Namespace = scheme + action Class Grant vs operation scheme and Class different_namespace 4 Path coverage (same namespace) Literal path segments, trailing-wildcard depth literal_mismatch, wildcard_requires_trailing_segment 5 Explicit empty bound Any grant/intersection source declares a parameter value that is []/{} (a present-but-empty params object is no constraint, Section 7 rule 6) empty_bound_denies_class 6 Null values Any null parameter value invalid_params_null 7 Param presence Grant bounds a param the operation omits (or operation has no params); operation carries a key the grant does not declare. A param_bounds key with optional:true is exempt from the omission half (Section 6.5) params_missing, undeclared_param 8 Enum membership and cardinality Request value not a member of a granted array set (Section 6.2); request cardinality outside a param_bounds min_items/max_items (Section 6.5) not_in_enum, params_cardinality 9 Bound comparison Numeric/bound exceeded; quantity value above the declared unit's bound; quantity unit differs from the declared unit (no conversion); param_bounds min/max out of range; step not an integer multiple (Section 6.5, Section 6.7) params_exceed_grant, unit_mismatch, params_out_of_range, params_not_multiple 10 Coverage emptiness Intersection result empty; zero sources; no grant covers the operation no_overlap, absent_source, capability_not_authorized 11 Constraint evaluation Non-recognized (scheme,type); recognized type value out of Section 8.1 grammar; recognized type violated unknown_constraint, invalid_constraint, {type}:violated Notes: Namespace is scheme:action_class (the first two :-delimited segments). Identifiers in the same namespace differ only below the Class; such mismatches are path-level (literal_mismatch / wildcard_requires_trailing_segment), not different_namespace. Params normalization fires at the input boundary: duplicate keys, over-precision/non-finite numbers, and size/depth overruns (Section 6.2) are detected before any grant-vs-operation comparison and therefore before every other layer of this table. Deny-when-declared: layer 5 is evaluated on the grant/intersection side — an explicitly empty [] or {} param value denies the class regardless of the request, before any member or bound check. params:{} ≡ absent: the params constraint looks only at the set of declared param names, independent of carrier form — an absent params and "params":{} both mean "no param names declared" = no param constraint (consistent across entailment and intersection; the same representation cannot have different semantics in different functions). Therefore: Direct grant: a grant with params:{} allows an op carrying any params (.{x:1} does not trigger undeclared_param); Intersection: a params:{} source contributes no param constraint ({limit:50} ∩ {} = {limit:50}); Key closure (next bullet) applies only when the grant declares a non-empty set of param names. Layer 7 key closure runs both directions: granted keys must be present in the operation (params_missing) and operation keys must be declared by the grant (undeclared_param); the missing-key check resolves first, and both precede the layer 8–9 value checks. (Applies only to a non-empty set of declared names, see the bullet above.) Layer 11 runs last by construction: constraints are evaluated only after coverage and parameters pass. Multi-grant aggregation (normative): Authorize operates on a ordered list of grants, not a single grant. Authorization semantics are order-independent; only reason selection (rule 4) uses the input order. Rules: Any one covering-and-allowing grant allows (∃ g: Entails(g,op) ∧ no params-layer rejection ∧ no constraint-layer rejection); Residual obligations = unresolved union across the covering grants that allow (normalized + sorted). A covering grant rejected at the params/constraint layer contributes none — it does not authorize, so its residuals are not carried (they are never silently dropped from a decision it does not produce); No grant covers → capability_not_authorized; op layer-1 errors always precede any coverage/aggregation decision; Some grant covers but all covering grants reject at the params/constraint layer → deny, reason = the layer 5–11 rejection reason of the first covering grant in canonical order (deterministic). Canonical order = the appearance order in the input grant list (the capability-record body order); an implementation MUST NOT choose the reason by internal hash/iteration order. Counter-examples pinned (forbidden): NOT any-one-allows — otherwise, with multiple grants held, a narrow grant would wrongly deny the legal operation of a broad grant; NOT first-match-wins — otherwise grant order would change the authorization outcome; NOT all-must-pass — synonymous with "any-one-covers authorizes", letting a narrow grant's existence invalidate a broad grant. Op-ID validation errors propagate their specific layer-1 code (missing_capability_id / unsupported_wildcard / invalid_capability_id), never capability_not_authorized and never an invented catch-all: step 1 reports the very code that describes the operation id. Only coverage failures (layers 3–4 and 10) collapse to capability_not_authorized. Layer 1 validates the operation id only. A malformed id in a grant makes that grant non-matching for every operation: Entails reports the specific layer-1 code as its false reason, and Authorize folds the non-match into coverage (capability_not_authorized) — the grant's own code is not surfaced by the decision. An absent/empty effective grant drops straight to layer 10 (capability_not_authorized); the evaluator MUST return this decision rather than raising. An absent operation drops to layer 1 (missing_capability_id). A language revision mismatch (Section 12.1) is resolved before any layer and reports unsupported_language_revision (fail-closed, no downgrade). Resolve (Section 8.5) is post-decision, not a layer. It consumes a Decision and never re-runs Authorize; only invalid_resolution / invalid_timestamp (Section 9.2) and a {type}:violated obligation result are introduced by it, and only on an allow_unresolved input.

Reason Codes (normative) Reason codes are stable identifiers. v1 defines: Canonical code = everything before the first :. An implementation MAY append : <detail> (e.g. the offending parameter name) as a diagnostic suffix; the canonical code is unchanged. All tooling and compatibility checks MUST compare the canonical prefix only. Reason code Meaning unsupported_wildcard Wildcard shape forbidden in v1 (bare *, partial segment, **, {a,b}, [a-z]) invalid_capability_id CapabilityId does not conform to the Section 3 grammar missing_capability_id Operation has no id (layer 1) invalid_params_duplicate_key Params contain a duplicate JSON key (Section 6.2 representation, Section 9.1 layer 2) invalid_params_number The params input has no canonical form: a numeric param is non-finite, out of IEEE-754 range, or over-precision (> 17 significant decimal digits) (Section 6.2 step 3), or the text is not well-formed Unicode / not parseable as a JSON object (Section 6.2 step 7). One stable code covers every "cannot be normalized" params failure (Section 6.2 representation, Section 9.1 layer 2) invalid_params_size Params exceed the 512-byte serialized size or the depth-32 nesting limit (Section 6.2 representation, Section 9.1 layer 2) invalid_params_binding A param_bounds bound is malformed (unknown member, mixed families, min > max, step ≤ 0, min_items > max_items), or a key is declared in both params and param_bounds (Section 6.5, Section 9.1 layer 2) different_namespace Grant and operation differ in namespace (scheme + action Class, Section 9.1 layer 3) literal_mismatch Literal identifiers differ wildcard_requires_trailing_segment Wildcard has no remaining segment (...:* does not cover ...) capability_not_authorized No grant in the effective set covers the operation no_overlap Intersection of multiple sources is empty absent_source Intersection over zero sources — no effective set (Section 7 rule 5) params_exceed_grant Request parameters exceed the granted bound params_missing Grant bounds a parameter but the request omits it, or the request has no params field at all (fail-closed, Section 6.3 step 4) undeclared_param Operation parameter key not declared by the grant's params (key closure, Section 6.2; Section 9.1 layer 7 request side) empty_bound_denies_class An explicitly empty bound at a parameter value ([]/{}) denies the class. params:{} is not an empty bound: it is equivalent to an absent params (Section 7 rule 6) not_in_enum Request value is not a member of the allowed set granted as an array (Section 6.2 enum rule) params_cardinality Request cardinality is outside a param_bounds min_items/max_items (Section 6.5) params_out_of_range Request value is outside a param_bounds min/max inclusive bound (Section 6.5) params_not_multiple Request value is not an integer multiple of a param_bounds step, per the IEEE-754 rule (Section 6.5) unit_mismatch A quantity value's unit differs from the unit fixed by the scheme declaration (Section 6.7). Units are never converted, so a different unit is never covered invalid_params_null null parameter value (rejected in v1) unsupported_language_revision Declared CLC revision is incompatible with the implementation (Section 12.1; fails closed, no silent downgrade) unsupported_domain The scheme declaration names a domain this implementation does not support, or the declaration is malformed / its digest does not match (Section 6.7, resolved before Section 9.1 layer 1) unknown_constraint Unknown constraint type (fail-closed) invalid_constraint Recognized type whose value fails its Section 8.1 value grammar {type}:violated A known constraint is violated (e.g. max_rows:violated); also reported by Resolve for a discharged-as-violated obligation (Section 8.5) invalid_resolution A Resolve resolution entry is malformed (unknown status, non-string/empty constraint) (Section 8.5) invalid_timestamp The Resolve now argument is not a valid UTC instant (Section 8.5) Other schemes MAY define additional codes, but MUST NOT redefine these.

Satisfaction Function (Evidence Side)

Satisfy(evidence_set, requirement) → Satisfaction Satisfaction = { verdict: "SATISFIED"|"UNSATISFIED", reason: string|null }

Algorithm: Verify each evidence artifact under its native rules. For each required evidence role, check that an artifact fills it. Check that each artifact is bound to the exact action via Match (Section 6.4). Evaluate freshness, consumption, and role constraints. The evidence-side value grammar is defined in this revision: varwof/evidence-v1:freshness:sec:<n>, :consumption:once, :quorum:distinct:<n>, :exclusion:initiator|executor (Section 8.2). A recognized constraint whose evaluation belongs to the enforcement point (consumption) evaluates to unknown, which at the top level yields UNSATISFIED (never SATISFIED) — the evidence side has no allow_unresolved; unresolved is an authorization-side channel (Section 8.4, Section 11). All required roles filled and bound → SATISFIED. Any role unfilled, unbound, or violated → UNSATISFIED. Tri-state evaluation, binary report. A recognized evidence-side constraint is evaluated three-valued (satisfied / violated / unknown). Satisfaction itself is binary. A constraint that evaluates to unknown at the top level — including one whose evaluation belongs to the enforcement point (consumption) — MUST produce UNSATISFIED with a stable reason, never SATISFIED. unknown is an internal evaluation result, not a third top-level verdict. Properties: deterministic, fail-closed, stable reason codes.

Semantic Boundary CLC-v1 defines the shared minimal vocabulary and evaluation algorithms for both authorization and evidence. A projection is not a third decision language. When a carrier profile maps native authorization or evidence artifacts into the objects defined here, it does not add a verdict source. Authorization and evidence remain distinct, but share the ActionId (Section 4.3) and constraint vocabulary (Section 8) needed to connect a proposed action with evidence about the action that occurred. A projection MUST NOT report a native check or relation as satisfied unless the carrier's verifier establishes it; carrier-specific fields, trust decisions and runtime state remain profile responsibilities. CLC-v1 does NOT define: Trust models: who signs what, issuer trust, delegation chains (belongs to AIC-JWT , OAuth, SPIFFE, etc.) Native verification: signature checking, schema validation, freshness enforcement (belongs to each native artifact's spec) Execution lifecycle: consumption, invocation, reconciliation, outcome classification (belongs to EMILIA AEB or equivalent) Receipt or token formats: the wire formats for carrying grants, evidence, or bindings (belongs to protocol-specific specs) The boundary is: CLC-v1 defines what to evaluate (grant ⊆ operation, evidence ↔ action) Consumers define how to evaluate (native verification, trust anchors) CLC-v1 defines what the output means (allow/deny/allow_unresolved, SATISFIED/UNSATISFIED) — allow and allow_unresolved are distinct enum values; a consumer MUST NOT treat allow_unresolved as allow (Section 8.4) Consumers define what to do with the output (invoke, record, reconcile) CLC-v1 defines each known type's value grammar (what counts as a legal constraint value); the declaring scheme defines how that value is evaluated (whether this window/CIDR currently forms a boundary) allow_unresolved is an authorization result, not evidence. It marks an unresolved authorization (or policy) condition. A consumer that can evaluate the obligation under a pinned rule may release it; one that cannot MUST refuse. It takes on an evidence role only where a relying party separately defines one together with the native verifier for it (AEB); the language itself makes no such claim, and unresolved MUST NOT be read as "evidence still required" A Section 6.2 refusal does not transfer to a decoded value. The input-boundary checks are defined over the text as received (Section 6.2). A deployment that must reproduce a refusal, or that applies a permit to a request it did not evaluate, therefore anchors the decision to the received octets or to a digest of them (Section 4.3) — never to a value a decoder has already normalized. A pipeline that can apply a permit to a request whose text was never checked has left the CLC boundary CLC-A stays the scope language: material-action identity is referenced from CAID , heterogeneous evidence evaluation from AEC , the boundary lifecycle from AEB , and durable consumption/accounting from BCR ; the narrow crosswalk between them is the composition point Principal/agent delegation and token attenuation stay with PAP and AAT ; CLC consumes their authorization outcome rather than redefining it A relation between two artifacts is not expressible. CLC evaluates a role expression over facts, each with its own constraint; it has no term for a relation that connects two artifacts — a carrier's "this permit is for that quorum" (draft-schrock-ep-authorization-evidence-chain-08 Section 4 required_bindings) or an audit record's derived agreement between a decision and an observed effect (draft-gilda-audit-agent-audit-record-00 Section 5.3). A carrier that carries such a relation MUST normalize it into a fact member that its own verifier establishes, or enforce it outside the projection; the presence of the two endpoints MUST NOT be read as the relation holding, and CLC MUST NOT report an unchecked relation as satisfied. A carrier that binds by identity instead (draft-thallapelly-oasnt-02 Section 6.2 adg) needs no such rule: identity is what Section 4.2 and Section 6.4 already evaluate

Conformance CLC-v1 defines three conformance classes: CLC-A (authorization side) — the v1 baseline. A conforming implementation MUST implement: grammar (Section 3), entailment (Section 6.1), intersection (Section 7), decision function (Section 9), rejection of non-recognized (scheme,type) constraints (unknown_constraint, Section 8.1), rejection of non-conforming recognized-type values (invalid_constraint, Section 8.1), exposure of recognized-but-unevaluated constraints via the decision's unresolved field on an independent allow_unresolved verdict (never silently dropped, Section 8.4), multi-grant aggregation (Section 9.1), and stable reason codes (Section 9.2). CLC-D (delegation side) — the containment relation, defined in Section 13. A conforming implementation MUST implement Contains(parent, child) with its ordered layers, the relation's two stable reason codes (child_exceeds_parent, params_not_narrower) and its profile contract (Section 13.8.1), and MUST pass containment-vectors.json. The third CLC-D code, delegation_mode_not_narrower, is produced by the binding profile's delegation-mode pre-check (Section 13.4.5) — never by Contains, which takes no mode argument — and belongs to the profile's obligations. A delegation policy that requires each hop to stay inside the previous hop's boundary reads Contains, not Section 7: entailment and intersection over the declared sets are necessary but not sufficient, because they do not compare a child's boundary against a parent's. An operation fitting a grant proves nothing about a child staying inside its parent, and where containment cannot be established a binding profile MUST NOT authorize the delegation. Contains is a language relation over grant values, not a wire format; the effective subset a delegation record carries is still the profile's to define. CLC-E (evidence side) — optional conformance profile, implemented and pinned by a corpus in this revision, but NOT claimed. Section 6.4 (match) and Section 10 (satisfaction) define the evidence-side relations; this revision also defines the evidence-side constraint value grammar (varwof/evidence-v1:freshness:sec:<n>, :consumption:once, :quorum:distinct:<n>, :exclusion:initiator|executor) and ships a reference implementation and a corpus. The class is withheld on principle, not for lack of material: agreement is the bar, and the bar is two independent implementations (Section 12 of the principles document, P12; Section 12 here, "Independence of implementations"). Parity between implementations that share an author does not meet it. A second precondition is stewardship: the evidence-side semantics are the subject of joint review with EMILIA, so no claim is made ahead of that review. A future revision that claims CLC-E would carry these obligations, and an implementation that claims it today MUST: evaluate the four evidence-side types over eligible evidence facts, returning a three-valued result (satisfied / violated / unknown) where unknown is never read as satisfied, and report a recognized constraint whose evaluation belongs to the enforcement point (consumption) as unknown rather than satisfied; take eligibility from integrity-protected native results only: a fact that did not reach VERIFIED, or whose protected subject identifier is absent, MUST NOT be counted for quorum or exclusion; implement instance identity and binding (Section 4.2/Section 6.4): an ActionId is the digest of the JCS canonical serialization of the declared material projection, an undeclared field MUST NOT affect it, a missing declared material field makes the action non-matchable (never inferred or defaulted), and a comparison across suites or action types is INDETERMINATE — a mapping problem, never a match — unless a relying-party-pinned Action-Mapping Profile projects it; implement CLC-REQUIREMENT-v1 as a closed object (an undefined member is rejected) whose expression uses the bounded grammar — AND/OR with equal binding strength, evaluated strictly left to right, parentheses as the only precedence mechanism — and treat an identifier with no eligible component as false; take the requirement from relying-party configuration: a requirement supplied by the presenter MUST NOT be accepted or weakened; pass evidence-vectors.json (Section 12 conformance corpora). An implementation that implements only CLC-A MUST NOT claim CLC-E. CLC-E does not add a wire format: carriers that need one (e.g. an Action Evidence Envelope) profile Section 6.4/Section 10 themselves. Conformance corpora. The suites below are published in the repository tree pinned as ; repository paths written as capability/... throughout this document are relative to that pinned tree, so the exact vectors named here are retrievable. CLC-A conformance is exercised by two machine-readable reference suites at capability/data/_vectors/clc-v1/: vectors.json — 154 vectors mapped to Appendix B — and property-cases.json — 1184 cases pinning the Section 7 meet-law, identifier narrowing and source-order independence. Their syntax is defined by vectors.schema.json; offline-vectors.json is a timestamped snapshot mirror. A conforming implementation MUST pass both suites. CLC-D conformance (Section 13) is exercised by containment-vectors.json — 64 vectors mapped to Section 13 — together with containment-property-cases.json (784 cases pinning the Section 13.3–Section 13.7 narrowing laws). containment-crosswalk-vectors.json carries 45 cross-vendor vectors that map six adjacent capability representations into CLC grants through pinned profiles and assert the containment verdict the unchanged relation reaches (Appendix C). Genericity is exercised, not asserted. crosswalk-vectors.json in the same directory carries 13 vectors in both directions: 5 that project a CLC decision into the members an AEB crossing record asks of a native source (the members CLC can establish, and the ones it explicitly does not), and 8 that map five foreign capability representations — OAuth RAR authorization_details, an AIC-JWT delegation authorization, an Action Evidence Graph capability class, a UCAN {with, can} capability, and a delegation chain — into CLC grants through pinned cross-walk profiles and assert the decision the unchanged core reaches. A profile is a few lines of mapping written by whoever owns the foreign format; the core is not modified for any of them. The evidence side ships evidence-vectors.json in the same directory — 32 vectors covering the four evidence-side constraint types, their value-grammar rejections, requirement-expression binding, the closed requirement object, ActionId computation (Section 4.2: declared material projection, undeclared fields excluded, missing material field non-matchable, suite-tagged identifiers) and Match verdicts (Section 6.4: MATCH / NOT_EQUIVALENT / INDETERMINATE). Its runner ships with the reference implementation (register), and its syntax mirrors vectors.json. Implementations MUST NOT: redefine semantics, accept v1-forbidden wildcards, or broaden bounds during canonicalization. Independence of implementations (honest scope). The three implementations named in this repository's README (Go, Python, TypeScript) are not independent evidence: they share an author, and their agreement is a regression test for the specification, not third-party validation. An independent implementation is invited; until one exists, the parity claim in this document is scoped to "same-author, three languages, one corpus". Reviewers SHOULD treat a single-author parity claim as evidence that the text is implementable, not that it has been independently interpreted. Participated in the cross-check vs. published here. The normative gaps that this revision closes were found by checking each rule against several same-author implementations written in different host languages, because a carrier-specific value model is exactly where a decision rule goes wrong — an integer that one runtime widens and another refuses, a string comparison that orders by UTF-16 code units instead of UTF-8 octets. Only the three ports named above are published, and they run the machine-readable corpora at their ordinary test entry points. This document asserts nothing about the behaviour of any implementation it does not publish: a further same-author port that took part in the cross-check carries no public conformance claim, and its results are not part of the parity claim above. Readers SHOULD evaluate the rules on their merits — each is anchored to a cited requirement — and not on the number of ports that happened to surface it. Experimental neighbours are not CLC. The WIT/WPT interop study in varwof/aic-jwt (wit-wpt-interop/) is an experimental research artifact that implements a different, wider wildcard surface (**, {a,b}, [a-z]) which this revision rejects as unsupported_wildcard (Section 9.2). It is not a CLC-A implementation and MUST NOT be cited as one; it exists to study WIT/WPT provisioning and carries its own EXPERIMENTAL banner.

Language Revision Every implementation declares a language revision CLC-<major>.<minor> — this document declares CLC-1.19. A capability input (grant, operation, or OCM) SHOULD carry the revision it was authored against; an input without a declared revision is treated as CLC-1.0. Compatible reading: an implementation MAY evaluate an input whose declared major equals its own AND whose declared minor is ≤ its own (so an implementation of CLC-1.3 reads a CLC-1.0/1.1/1.2/1.3 input, but not CLC-1.4 or CLC-2.0). CLC-1.2 is additive: it adds the unresolved field to the Decision shape and the invalid_constraint reason code without changing v1 verdicts on existing inputs; a CLC-1.1 implementation MAY claim CLC-1.1 against this document but is not CLC-A conformant (Section 12) until it exposes unresolved and rejects non-conforming recognized-type values. CLC-1.3 is additive, with one verdict value re-scoped: it adds the allow_unresolved verdict value, closes the decision loop for residual obligations, and re-scopes allow to mean "fully enforced" only; allow/deny outputs on inputs with no residual obligations do not change. A CLC-1.2 implementation MAY claim CLC-1.2 against this document but is not CLC-A conformant (Section 12) until it emits the allow_unresolved value and applies the Section 8.1 (scheme,type) identity and cross-midnight window grammar. CLC-1.9 is additive: it adds the containment relation and conformance class CLC-D (Section 13) without changing any CLC-A verdict, reason code or vector. A CLC-1.8 implementation that does not claim CLC-D MAY claim CLC-1.8 against this document and remains CLC-A conformant; an implementation that claims CLC-D MUST pass containment-vectors.json (Section 12). CLC-1.10 is additive with a minor gate: it adds the optional param_bounds field and its four reason codes (Section 6.5). Every input that declares no param_bounds is unchanged, so a CLC-1.10 implementation reads every CLC-1.x input. An input that uses param_bounds MUST declare CLC-1.10 or later; an implementation that does not implement Section 6.5 MUST refuse such an input through the minor gate (unsupported_language_revision) and MUST NOT ignore the field — silently dropping bounds would grant more than the input declares (fail-closed). CLC-1.11 is additive: it adds the Resolve function (Section 8.5) and its two input-error reason codes (invalid_resolution, invalid_timestamp) without changing any Authorize verdict, reason code or vector. A CLC-1.10 implementation that does not implement Resolve MAY claim CLC-1.10 against this document and remains CLC-A conformant; Resolve is exercised only when a consumer feeds a decision back, so no input shape is reinterpreted. A time:window obligation is still delivered as allow_unresolved (Section 8.4) — the core clock evaluator of Section 8.5 runs only inside Resolve with a supplied now. CLC-1.12 is additive: it adds the derived function ConstraintUnion (Section 7.1) without changing any other relation, verdict, reason code or vector. The function is exactly the constraint projection of Intersect rule 3, so Intersect's output is unchanged; an implementation that does not implement ConstraintUnion MAY claim CLC-1.11 against this document and remains CLC-A conformant. CLC-1.13 is additive and CLC-D-scoped: it adds AuthorizeWithChain (Section 13.11), a CLC-D function that fixes the order of Contains and Authorize over an effective-chain Intersect. It changes no CLC-A verdict, reason code or vector; an implementation claiming only CLC-A is unaffected, and a CLC-D implementation adds it to its CLC-1.9-era Contains surface. CLC-1.14 is additive: it defines the intersection of param_bounds (BoundMeet, Section 6.6) so that Intersect can combine a grant carrying the bounds Section 6.5 defines. It changes no verdict, reason code or vector for an input without param_bounds; it changes the refusal of an input with param_bounds from invalid_params_binding to the correct meet or empty-meet result. An implementation that does not implement Section 6.6 MUST refuse a param_bounds input through the minor gate rather than intersect it wrongly. Honest scope of that change (noted in rev CLC-1.15). One subset of inputs does not evaluate identically across the CLC-1.10→1.14 minor range: the inputs that declare param_bounds and reach a relation that meets them — Intersect (Section 7) or AuthorizeWithChain step 3 (Section 13.11) — with two or more sources declaring a Bound for the same key. On that subset the direction of the change is deny → allow: what CLC-1.10–1.13 uniformly refused (invalid_params_binding) becomes the correct meet result (usually an allow-side effective grant); where the meet is empty, the denial's reason changes from invalid_params_binding to no_overlap. Entailment against a single grant (Entails/Authorize), containment, and every input without param_bounds are verdict-stable across the whole 1.x range. The compatible-reading rule above is therefore intact — it governs readability, not verdict stability — but verdict stability across minor revisions does not hold for that one subset, and a consumer MUST NOT assume it. (Rev CLC-1.15 moves the same subset again, in the opposite direction, for cross-family meets only; see its entry below.) CLC-1.15 is a corrective revision (review-driven; see the Revision History): the cross-family numeric ∩ enum meet exception of CLC-1.14 is removed, so every numeric × enum meet refuses with invalid_params_binding in either source order (Section 6.6); enum membership equal is defined as JSON type-sensitive equality (Section 6.5 layer 8); ConstraintUnion's deterministic ordering is pinned to UTF-8 byte order (Section 7.1); delegation_mode_not_narrower is attributed to the binding profile's mode pre-check and moved out of the core reason commitments (Section 13.4.5, Section 13.5, Section 13.6, Section 13.11); Contains antisymmetry is stated on semantic equivalence classes (Section 13.3); and AuthorizeWithChain states the caller's complete-authenticated-root-first-chain obligation (Section 13.11). It changes no verdict, reason code or vector for an input without param_bounds. On the Section 6.6 meet subset the direction partly reverses CLC-1.14: a cross-family numeric × enum intersection that CLC-1.14 reduced to a filtered enum — an allow-side effective grant — now refuses (allow → deny, invalid_params_binding), because the filtered enum was broader than either source (Section 6.6); same-family meets are unchanged. The same review adjudicated the scalar × nested pair (numeric or enum vs. object, either order) to the same cross-family refusal: that pair was already denied in CLC-1.14 (as no_overlap), so the direction there changes the reason code only (→ invalid_params_binding), not the verdict (Section 6.6, design-notes D12). The Section 8.4 residual-obligation list and the Resolve output re-use the Section 7.1 UTF-8 byte-order collation; an implementation whose native default ordering is UTF-16 code-unit order MUST apply the pinned comparison explicitly. An implementation that does not apply this revision MUST declare CLC-1.14 or earlier and let the minor gate (Section 12.1) resolve any input that relies on the corrected behavior. CLC-1.16 is a corrective revision (cross-port cross-check; see the Revision History): the I-JSON integer bound (Section 6.2 step 3), leading-BOM refusal on the raw path (step 8), host numeric-carrier normalization (step 9), the hardened decoded-entry obligation (step 7), the max_rows × key-closure reconciliation (Section 8.1) and the containment clarifications (Section 13.4.2 segments-are-not-a-hierarchy, Section 13.4.3 decision table). Honest scope of the verdict changes. Two input subsets move against CLC-1.15: (1) an integer-valued numeric param whose magnitude exceeds 2^53 − 1 was previously canonicalized to binary64 (rounded) and compared — it is now refused (invalid_params_number), so whatever verdict the rounded comparison produced becomes a deny; (2) raw params text with a leading UTF-8 BOM was previously implementation-split (an ECMAScript-whitespace trim accepted it) — it is now uniformly refused, an allow → deny for the trimming implementations. The Section 8.1 and Section 13.4 additions codify existing behavior and change no verdict; every other well-formed input is verdict-stable against CLC-1.15. An implementation that does not apply this revision MUST declare CLC-1.15 or earlier and let the minor gate resolve an input that relies on the corrected boundaries. CLC-1.17 is a corrective revision (numeric-corner closeout; see the Revision History): Section 6.2 step 3 is restated so the I-JSON integer bound is judged on the mathematical value of the received token — an integer-valued number outside [−(2^53 − 1), 2^53 − 1] is refused whether or not binary64 represents it exactly (9007199254740992 is exactly representable but refused; 1e100 is not exactly representable and is also refused), correcting a CLC-1.16 justification that had said such a number "has no exact IEEE-754 binary64 representation". Step 3 additionally states that number identity is the JSON number, not its spelling (-0 ≡ 0, 1e3 ≡ 1000, 1.5e-3 ≡ 0.0015) and pins a non-integer token of ≤ 17 significant digits to its IEEE-754 binary64 value. Honest scope: no verdict changes. Every clause restates behavior CLC-1.16 already introduced (Section 6.2 steps 1 and 3), so no verdict, reason code or vector moves for any input a CLC-1.16 implementation evaluates; the revision exists because the restatement is normative text and the corpus gains the corner pins. Corpus: 146 → 154 vectors (params-063–params-070, kind=entail, tagged CLC-1.17) and param-bounds-vectors.json 43 → 49 (pb-044–pb-049). The citation is corrected to Section 2.2. CLC-1.18 is additive (binding-profile contract completion; see the Revision History): Section 13.8.1 gains two profile obligations — a profile that performs a chain check MUST declare its transitivity basis (definition-derived, mechanically-checkable or asserted, with the finite domain and deterministic procedure stated for the second), and a profile MUST NOT report a union axis as a containment result. New Section 13.9.7 is informative and records how a carrier's own decision states map onto the three CLC verdicts and the Resolve values. Section 11 adds the evidence-side counterpart: a relation between two artifacts is not expressible, so a carrier either normalizes it into a fact that its own verifier establishes or enforces it outside the projection, and the presence of both endpoints is never the relation. The profile obligations are on profiles, not on the core relation: Contains remains a per-pair predicate that claims no transitive closure and reads no constraint, so no CLC-A verdict, reason code or vector changes and a CLC-1.17 implementation stays CLC-A conformant (it loses only the ability to claim the new profile obligations for a profile it publishes). A carrier profile written against Section 13.8.1 before this revision MUST either add the two declarations or state that it is not claiming a chain check. CLC-1.19 is additive with a minor gate: it adds the scheme declaration (Section 6.7) and four declared domains (quantity, decimal-string, subdomain, path-prefix), plus the two reason codes they need (unit_mismatch, unsupported_domain). A parameter with no declaration keeps the shape-inferred domain, so every CLC-1.18 input is unchanged; an input that relies on a declared domain MUST declare CLC-1.19 or later, and an implementation that does not implement Section 6.7 refuses it through the minor gate (unsupported_language_revision) rather than evaluating it under the shape rule. No CLC-A verdict, reason code or vector changes for an input without a declaration. CLC-A conformance and the minor gate are the two sides of one rule. Claiming CLC-A (this section) means implementing the CLC-A-relevant semantics of the revision claimed — so an implementation that advertises CLC-1.19 MUST implement param_bounds (grammar and the Section 6.6 meet), Resolve, ConstraintUnion and the Section 6.2 canonicalization, not merely tolerate their inputs. The minor gate is the complement for an implementation that lags: it declares an older revision and refuses any input that uses a field or function introduced after it. The two MUST agree — an implementation MUST NOT claim a revision higher than it implements (which would silently evaluate newer inputs), nor claim CLC-A while refusing a well-formed input of its own declared revision. Incompatible reading MUST fail closed with deny("unsupported_language_revision"). An implementation MUST NOT silently evaluate under a different revision — no downgrade, no warning-then-allow. The revision check resolves before any Section 9.1 layer and yields the single resolved reason code unsupported_language_revision. Vectors: revision-001 (input CLC-1.0 against an implementation declaring CLC-1.3 → eval normally, allow); revision-002 (input CLC-2.0 → deny unsupported_language_revision).

Delegation Containment This section defines containment, the third grant-level relation of CLC-v1 alongside entailment (Section 6.1) and intersection (Section 7), and the conformance class CLC-D that exercises it. It is additive: it changes no CLC-A verdict, reason code or vector, and a CLC-A implementation that does not claim CLC-D is unaffected. It was folded into this document from the formerly separate containment extension, which is retired.

Motivation The two relations the core defines answer different questions: Entailment (Section 6.1): does an operation fit inside a grant? Intersection (Section 7): what is the effective set when several grants cover one authority? Neither answers the question a delegation chain asks of every hop: is the child's declared grant inside the parent's declared grant? Intersection delivers a set that is necessarily common to the sources, but a result fitted to two grants proves nothing that one of the grants' boundaries would not also authorize on its own — it reasons about the combined set, not about a child that must be a subset of a specific parent. The AIC ecosystem needs the child-parent question at two concrete boundaries: Delegation (DelegationAuthTBS vs. a principal's grant): a sub-agent's requested capabilities, constraints and delegation mode must lie inside what the principal authorized. Without a shared relation this obligation is the profile's job and is re-implemented per profile, so two profiles can differ on the same sub-agent grant. Publication bound (rule signing): a rule a certificate holder publishes must not exceed the holder's own grant. This section states the general relation; register/ruleexec already applies the same idea in one instance (RuleWithinSignerGrant), and is a candidate early adopter. Containment is intentionally additive, not a CLC-v2 feature: it does not change any CLC-A verdict, does not touch the CLC-A corpus, and can be adopted by any implementation that claims CLC-A without breaking compatible reading of existing inputs.

Terminology and Notation Grant and Operation are as defined in Section 2/Section 5 (identifier per Section 3, parameters per Section 6.2, constraints per Section 8). Contains(GP, GC) names the relation parent GP × child GC. Declared set — the parameters a grant carries as constraints on operations (Section 6.2 value semantics: numbers bind as upper bounds, arrays as membership sets, objects recursively per key, explicit empty []/{} deny the class, {}≡absent). Bound — a declared constraint value as interpreted by Section 6.2/Section 8.1 value semantics. Narrower — a child grant is narrower than its parent when every value it declares is within the parent's declared bounds and its key set is closed by the parent's. Where the carrier defines a delegation mode, the child's mode must not widen the parent's either — that half is the binding profile's pre-check, not part of the language relation (Section 13.4.5). Mode lattice — an abstract carrier-defined order over delegation modes, exercised by the binding profile's pre-check (Section 13.4.5), never by Contains; the AIC-JWT ordering is pinned in Section 13.8.

Relation Signature

Contains(GP: Grant, GC: Grant) -> ContainmentResult ContainmentResult = { // JSON object (not ASN.1) "contains": <boolean>, // false <=> one of the Section 13.4 layers failed "reason": <string> // resolved reason code (core Section 9.2 code) }

contains is true only when every layer of Section 13.4 passes. reason on success is empty; on failure it carries the first failing layer's reason code, per Section 13.4 layer order (deterministic: input order of the two grants never influences which layer reports first). The relation is antisymmetric on semantic equivalence classes, not on Grant objects: Contains(A,B) and Contains(B,A) both hold exactly when A and B fall in the same class — they denote the same identifier coverage, the same declared parameter key set and the same bounds under the Section 6.2/Section 6.5 value semantics, with surface-equivalent spellings identified (params:{} ≡ absent, Section 6.2). Two syntactically different Grant objects in one class (an equivalence, not an identity of JSON text) therefore contain each other; containment of grants in two distinct classes in both directions is a contradiction and MUST NOT be reported. Delegation modes are not part of the relation (Section 13.4.5), so mode equality is neither required nor observable here — where a carrier binds modes, the profile's pre-check owns that comparison.

Containment Algorithm The relation resolves layers strictly in order; the first failure determines the reason code. Every layer is fail-closed: any doubt yields false.

Layer 1: Grant validity If a grant identifier is malformed per Section 3, Contains returns false with the matching CLC-A syntax code (invalid_capability_id / missing_capability_id). Validation errors are reused from the core, not re-invented. If GC.Params violates the grant-side parameter grammar (Section 6.2), the child is not a valid grant and Contains yields false (invalid_params_<n> codes as in the core). An invalid parent grant also yields false: a boundary that cannot itself be evaluated must not authorize a child.

Layer 2: Identifier coverage The child identifier must be covered by the parent identifier using exactly the CLC-v1 path-coverage relation (the rules Entails applies to identifiers, Section 6.1/Section 6.3, with parameters excluded — the identical keyset empty-object edge is not relevant here because parameters are excluded from this layer by construction): same namespace (scheme:action-class) — else different_namespace; same segment depth with equal literal segments, OR parent trailing * wildcard covering the child's trailing segments ("*" matches one or more segments, never zero) — else child_exceeds_parent. Mid-identifier wildcards remain unsupported_wildcard per Section 3: this section does not enlarge the v1 wildcard surface. Segments are not a hierarchy. The :-separated segments of an identifier read like a path, but coverage here is only segment equality or the trailing wildcard — a prefix relation between segments confers nothing. Contains(std/database-v1:query, std/database-v1:query:SELECT) fails at this layer with child_exceeds_parent (the child is deeper and the parent carries no wildcard), and the reverse order fails the same way: two identifiers of different segment depth cover each other only through a wildcard. The relation between an action class and an action inside it is established outside this relation — by whatever declares action relations (an action type definition) — never inferred from the identifier text. A caller building a delegation chain obtains coverage only from an explicit parent trailing wildcard (std/database-v1:query:* covers std/database-v1:query:SELECT), never from a shared prefix.

Layer 3: Parameter narrowing Every parameter the child declares must be within the parent's declared bounds, using the Section 6.2 value-subset semantics, and the key sets must be identical (symmetric closure). The child and parent key sets are each the union of params and param_bounds keys (Section 6.5): number (from params): child value ≤ parent bound (upper bound only in params; richer bounds are expressed in param_bounds, below); string / boolean: exact equality with the parent's value; array (enum): every child element must equal a member of the parent's set; a child empty array inside a non-empty parent enum is vacuously within it (child denotes "nothing", which is inside anything) — but a parent empty set denies the class (deny-when-declared), and a child empty set under a parent empty set is therefore false (params_not_narrower): the class is denied, not "narrowed to nothing"; object: recursion per shared key, with the same symmetric key closure at every depth (a child object may not add a key the parent object does not declare, and may not omit one the parent declares); param_bounds bound (Section 6.5): for a key with a Bound on the parent's side, the child's Bound for that key must be within it — a numeric family bound must have min raised or equal and max lowered or equal (min_child ≥ min_parent / max_child ≤ max_parent, and the child MUST declare a min/max the parent declares), a step must be an integer multiple of the parent's step (a coarser-or-equal grid whose values are a subset of the parent's allowed values), an enum family must be a subset with min_items/max_items tightened or equal, and a nested bound recurses. A parent key that is required (optional absent or false) forces the child's key to be required; a child MAY keep an optional key optional or make it required, but MUST NOT turn a required parent key optional. A child Bound that adds a family the parent does not declare, or omits a bound the parent declares, is params_not_narrower (the child would allow a value the parent denies); key closure: the child key set MUST equal the parent key set, both ways. A child that omits a parent-declared key would allow operations the parent denies (the missing key is params_missing in entailment); a child that adds a key the parent does not declare allows operations the parent denies (undeclared_param). Either failure is params_not_narrower. This is the same symmetric closure Section 6.2 layer 7 applies to an operation vs. a grant, carried over to grant-vs-grant comparison; extra rule: the parent's {} (or absent) params object is unconstrained and contains any child params; a child {} under a bounded parent is false (params_not_narrower) — declaring nothing is not the same as declaring a subset of the parent's bounds. Decision procedure. The rules above reduce to one table; presence is decided first (Section 6.5's four layers), then the key sets, then the per-key values. The table is normative over any prose reading: parent params/param_bounds child layer-3 result absent or {} absent or {} contains — decided by the identifier layer alone absent or {} present contains — an unconstrained parent contains any child params present absent or {} params_not_narrower — declaring nothing is not a subset of declared bounds present present, child key set equal to the parent's contains iff every child value is within the parent's value for that key (rules above) present present, key sets differing in either direction params_not_narrower — a key the child omits would allow operations the parent denies (params_missing in entailment); a key the child adds would allow operations the parent denies (undeclared_param) Both shorthand readings of "narrower" are wrong and are refused here: fewer keys is not narrower (a child that omits a parent-declared key lifts the parent's bound on it), and more keys is not narrower either (an extra declared key looks more constrained on the child's own operations, but containment asks whether every operation the child authorizes is one the parent authorizes — and an operation carrying the child's added key is undeclared_param at the parent). Key-set equality (the symmetric closure above) is the only shape that passes layer 3, decided before any per-key value comparison. The presence semantics match Section 6.2 exactly: a grant (parent or child) whose params are present-but-empty {} is unconstrained, identical to an absent params object (Section 9.2/Section 6.2). Where a scheme declaration (Section 6.7) names a domain for a key, the per-key value comparison is made under the parent's declared domain: quantity compares normalized values of the same unit, decimal-string compares value spaces, subdomain compares label suffixes and path-prefix compares segment boundaries. If the two sides declare different domains for the same key, the pair is not comparable and layer 3 fails closed with params_not_narrower — a child is never narrower merely because it states a different domain.

Constraints are outside the relation Constraints are not part of Contains. A grant carries constraints (Section 8), but they are a separate axis with a different composition rule: Constraints compose by union (conjunction), not by subset. Along a delegation chain the effective constraint set is the union of every link's constraints (Section 7 Intersect already unions them). A child therefore need not re-declare its parent's constraints, and adding or tightening a constraint only narrows. Contains compares identifier and parameters only. It does not read, compare, or validate the constraints field: a constraint difference never changes the Contains verdict. Constraint grammar and evaluation stay where they already are. Whether a constraint identity is recognized, whether its value is in grammar, and whether an operation satisfies it are the CLC-A concern of the consumer and Intersect/the decision function (including the allow_unresolved residual-obligation channel, Section 8.4). This section neither re-decides nor weakens them. Consequences a reviewer should take as intended: Contains(P, C) does not by itself assert that C's operation set is inside P's when constraints are in play — the parent's constraints are carried forward by the chain's union (Intersect), and a consumer that uses Contains as its only gate must compose the chain's intersections as well (Section 13.8). This is the honest reading: containment is a relation over the declared identifier and parameter boundary, and constraints are enforced by the union, not by this relation.

Delegation-mode lattice (binding-profile pre-check) Where a carrier defines a delegation mode, the child's mode must not widen the parent's. This check is a required binding-profile pre-check, not a layer of the language relation: as with constraints (Section 13.4.4), mode is a carrier-level concept — Grant values carry no mode, and the relation Contains(GP, GC) takes no mode argument, so a core Contains verdict can never express a mode decision. A binding profile (Section 13.8.1) that maps a mode-carrying carrier MUST run the mode-lattice check itself, before or alongside each Contains call, and MUST report delegation_mode_not_narrower from the profile when the child's mode widens the parent's; it MUST NOT rely on Contains for that check and MUST NOT present a Contains verdict as evidence that the mode narrowed. The lattice order is carrier-defined (CLC-D does not invent modes); the AIC-JWT order is authorized < representative — a child may be authorized under a representative parent, never the reverse. A carrier without a mode concept has no pre-check to run.

Reason Codes CLC-D registers exactly three child-level reason codes; two of them are returned by the relation, and everything else a Contains result carries reuses CLC-A codes. The third is produced only by the binding-profile delegation-mode pre-check (Section 13.4.5): the relation takes no mode argument and MUST NOT return it. An implementation MAY collapse child_exceeds_parent for identifier failures into the core's capability_not_authorized at a boundary that must not reveal policy shape (Section 11), but MUST NOT collapse params_not_narrower; a profile that maps a mode-carrying carrier MUST NOT collapse delegation_mode_not_narrower either. Code Produced by Meaning child_exceeds_parent Contains, layer 2 (Section 13.4.2) child identifier not covered by parent identifier params_not_narrower Contains, layer 3 (Section 13.4.3) a child parameter is not within the parent's declared bounds / key set delegation_mode_not_narrower binding-profile pre-check (Section 13.4.5) — never by Contains child delegation mode (a carrier concept) widens the parent's There is deliberately no constraint reason code: constraints are not part of the relation (Section 13.4.4).

Conformance Class CLC-D CLC-D is an optional conformance class stacked on CLC-A. A conforming implementation: MUST implement Contains (Section 13.4) and the relation's two Section 13.5 reason codes, and pass the CLC-D corpus (Section 13.7); where the implementation also ships a binding profile for a mode-carrying carrier, that profile owns the delegation_mode_not_narrower pre-check (Section 13.4.5); MUST (rev CLC-1.13) implement AuthorizeWithChain (Section 13.11) and pass the authorize-chain-vectors.json corpus, so the one-call chain check is exercised by the same class that owns containment; MUST NOT alter any CLC-A verdict, reason code, or the CLC-A corpus — this section is strictly additive; MUST treat containment as declared-set comparison: it declares no consequence about execution lifecycle, time windows after the fact, or post-hoc bounds; where a binding profile cannot establish containment at delegation time, the profile MUST NOT authorize. An implementation claiming CLC-A does not claim CLC-D. A delegation profile (a binding profile (Section 13.8), an AIC-JWT DA validator, a certificate-issuance stack) that needs the boundary check SHOULD require CLC-D of the module it delegates to, and MUST NOT substitute intersection (Section 7) for containment. Implementation status (single-author parity). Contains is implemented in all three reference implementations (Go: register/semantics.Contains; Python: aic-capability-demo/clc_semantics.py contains; TypeScript: ts/clc_semantics.ts contains) and mirrors the cases this section pins. Agreement between implementations that share an author is a regression test for this text, not independent validation — the Section 12 honesty rule applies unchanged to CLC-D (see Section 13.7).

Corpus Because CLC-D is a new relation, its corpus is new and independent of the CLC-A suites (vectors.json / property-cases.json / crosswalk-vectors.json / evidence-vectors.json). The corpus ships as capability/data/_vectors/clc-d/containment-vectors.json (64 vectors) and a forward-closure property file capability/data/_vectors/clc-d/containment-property-cases.json (784 cases × 39 shared operations, generated by capability/scripts/gen-contain-property-cases.py), plus a cross-walk corpus capability/data/_vectors/clc-d/containment-crosswalk-vectors.json (45 vectors) that maps a carrier's native representation (AIC-JWT DA, OAuth RAR, UCAN, delegation chain, and the adjacent agent drafts ATN, AAT, AIP, AAE, AOA, AEGIS) to a grant on each side and asserts Contains (Section 13.9.3). Rev CLC-1.13 adds capability/data/_vectors/clc-d/authorize-chain-vectors.json (15 vectors) pinning the fused AuthorizeWithChain relation (Section 13.11). The vectors cover: identifier coverage (layer 2): trailing-wildcard coverage, namespace mismatch, depth mismatch, same-length literal mismatch, unsupported_wildcard kept stable, wildcard-requires-a-trailing-segment; parameter narrowing (layer 3): number upper bound at/under/over, enum subset / element-absent / scalar-member, empty child enum under non-empty parent, parent empty set deny-class, key-closure violation in both directions (child omits, child adds), nested object recursion and nested key closure, child {} under bounded parent, child null (layer-1 grammar), unconstrained parent {}; constraint non-participation (Section 13.4.4): a tighter, wider, added, dropped, or differently-identified child constraint, and a child constraint the parent lacks — every one MUST leave the verdict unchanged (all assert contains); validity (layer 1): malformed parent/child identifiers, null params; symmetry: both-directions containment identity, antisymmetry probe pair (contain-040/041), a fully-narrower combined grant. The delegation-mode lattice is not part of the language-level corpus: the relation Contains(GP, GC) takes no mode, so a profile that binds a carrier mode (the AIC-JWT DA validator, Section 13.8) exercises it itself. The language corpus therefore ships no mode vectors; a carrier adopting CLC-D MUST add mode vectors to its own profile corpus. The property the property corpus checks is forward closure: for every (parent P, child C) and every operation o in the shared sample, Contains(P, C) MUST NOT raise, and Contains(P, C) ∧ Entails(C, o) ⟹ Entails(P, o). This is the declared-set consequence that makes a delegation boundary sound, and the reason the Section 13.5 collapse of layer-2 failures into child_exceeds_parent cannot weaken the boundary. On the shipped corpus the Go and Python/TS runs report the same 120 contained pairs and 30576 operation checks with zero violations. The parity bar and the "independence of implementations" honesty rule of Section 12 apply unchanged to CLC-D: until two independent implementations agree, the corpus is evidence that this text is implementable, not that it has been independently interpreted.

AIC-JWT / AIC Certificate Binding This subsection is a cross-walk profile of the relation, not part of the language. It covers the X.509 AIC model and its JWT representation . At the AIC delegation boundary the parent grant is produced from the principal's authorization and the child grant from the DelegationAuthTBS/AIC-JWT DA capabilities: Parent grant GP: the principal's capability entry (scheme:id params), plus principal-level authorizationConstraints projected as parent constraints, plus the principal's own delegation mode. Child grant GC: the sub-agent's requested capability (Capability.SchemeId:CapabilityId, Parameters), its authorizationConstraints, and its DelegationMode. Boundary result: P_effective = P_principal ∩ C_agent ∩ P_gateway keeps its intersection meaning — intersection determines the effective set; containment (Section 13.4) is the per-child admission predicate that runs before the intersection is composed, on each (parent-capability, child-capability) pair. The child's constraints are not compared by containment; they are carried forward by the intersection's union (Section 13.4.4), so the principal's constraints remain in force in P_effective. A sub-agent that requests a capability the principal did not grant fails at layer 2 (child_exceeds_parent); a sub-agent whose requested params exceed the principal's declared bounds fails at layer 3; a sub-agent that widens the carrier's delegation mode fails the lattice (Section 13.4.5). Binding profiles MUST surface the reason code into their audit trail, and MUST compose the intersection so the principal's constraints stay in force.

Profile contract (for any binding profile) A binding profile maps a carrier's native authorization structure to the CLC grants Contains compares (Section 13.9.3). The profile is the carrier's, not the language's (a carrier's field names appear in its own profile, never in the language text). Because every Section 13.9.3 mapping is a profile, this subsection states the obligations a conforming profile carries. It adds no CLC-A or CLC-D verdict and changes no relation. Resolve the carrier's own inheritance and defaults before mapping. The language has no inheritance, no default values and no "absent means inherit" rule (Section 13.10.1). A carrier that resolves an absent dimension against an ancestor (AIP Section 4.4) MUST do so in the profile and emit the resolved declared set. Contains is defined over the grants the profile emits, so an unresolved default is a profile defect, not a containment result. Preserve every identity dimension the carrier treats as identity. If the carrier treats two artifacts with the same name but different metadata as distinct capabilities (ATN Section 9.1: same id, different schema.digest), the profile MUST carry that metadata into the mapped grant — here, a trailing identifier segment — so a mismatch fails closed. A profile MUST NOT drop an identity dimension and then compare the artifacts as equal. Residualize semantics the relation cannot see; never drop them silently. Dimensions the mapped grant has no field for (AEGIS allowed_roles, environment, risk_level; AAE validity) are not compared by Contains. The profile either models them in its own document and enforces them outside the relation, or declares them out of scope — but it MUST NOT report containment as if they had been checked (fail-closed: what the relation cannot see, it does not permit). Do not invent carrier semantics the carrier does not state. A profile must not widen what the carrier leaves undefined into an allow. AEGIS's dotted ids are hierarchical, but AEGIS grants capabilities individually and does not define domain-level containment, so the profile does not turn a bare domain into a namespace wildcard (ccx-042 fails closed). Non-goals (recorded, not prohibitions on carriers). Three properties are deliberately outside both the relation and the profile contract: union of authority sources — a sub-agent that combines narrow delegated authority with broad independent authority (AEGIS Section 5.1, AOA) composes over a set of grants, which is carrier composition governance, not a per-pair predicate; chain-level verification — Contains is a per-pair predicate, not a transitive closure; a chain check is the profile's iteration of Contains over hops (Section 13.9.3, delegation-chain->clc-v1); cross-carrier identity — CLC defines no equivalence between two carriers' capability names; a profile MAY publish its own mapping, but the language asserts none. Declare the transitivity basis before claiming a chain check. Contains is a per-pair predicate: it decides one adjacent-to-adjacent pair and asserts nothing about a chain, a closure, or a composed authority (Section 13.3, Section 13.9.3). A profile that checks a chain of more than one hop is therefore making a claim the language does not make for it. Such a profile MUST state its transitivity basis as exactly one of: definition-derived — the relation is transitive by construction for the carrier's identifier grammar and parameter family, with the argument recorded; or mechanically-checkable — the profile declares the finite domain the claim ranges over and the deterministic procedure that decides it, so a third party can re-run the claim; or asserted — the carrier states the property and no procedure is offered. The basis is a property of the profile's chain claim, not of Contains; a profile MUST NOT describe a chain check as a CLC relation, and a consumer MUST NOT read Contains(h1, h2) plus Contains(h2, h3) as entailing Contains(h1, h3) without the profile's declared basis. A deployment that needs only the pairwise judgement (every hop inside its own parent) needs no basis and MUST NOT claim one. This distinction is also applied to inter-domain agent delegation chains in Section 8: per-hop comparisons support a chain-wide statement only when the shared intermediate value is authenticated and all hops use the same immutable profile and basis. That requirement is informative and routing-layer-specific here; it neither changes Contains nor supplies the authenticated adjacency it requires. Never report a union axis as a containment result. A carrier field that is conjunctive over a chain — accumulated constraints, ATN preconditions, AAE CONSTRAINTS, a requires precondition list — is a union axis: each hop's burden adds to the chain's, and the composition operator for it is the constraint union of Section 7.1, not the relation of Section 13.4. A profile MUST route such a field to the union function (or to its own enforcement outside the relation) and MUST NOT report it as contained, narrowed, or checked-by-Contains. The failure mode this obligation closes is a chain reported as contained while a hop's constraint was neither compared nor carried — a null constraint check is not evidence of containment (Section 13.4.4). Evidence. Both obligations are pinned by data rather than by prose: data/_vectors/clc-v1/profile-declarations.json carries one declaration per profile used by the two cross-walk corpora, together with the illustrative declarations that pin the accept path and the negative cases that pin each refusal (basis-missing, domain-or-procedure-missing, basis-without-claim, unknown-basis, union-axis-misrouted, union-axis-fields-mismatch). The declarations are checked — including the requirement that the declared profiles equal the profiles the corpora actually use — by the Go runner (register/internal/crosswalkvectors, TestProfileDeclarations) and by its Java counterpart (com.varwof.clc.runner.ProfileDeclarationsTest). Both implementations share an author, so this is regression evidence for the rule, not the independent-implementation bar of Section 12.

Related Work and Positioning This subsection is informative and carries no requirements. It records why a language-level containment relation is needed at all, how CLC relates to the authorization work already in flight, and what the adoption risks are.

The gap CLC fills CLC does not define a carrier — it is not a credential format, a transport, or a policy store. It defines the evaluation semantics that carrier-facing work leaves open: a deterministic decision, stable reason codes, and (here) a containment relation. The recurring shortfall in today's landscape is not "how do I carry an authorization" but "how do two independent implementations agree on what a carried authorization means, and on whether a delegated one stayed inside its parent".

What containment is for A standardizable evaluation core. Structured authorization payloads exist, but their semantics are pinned by each profile's own type vocabulary, so cross-implementation agreement is not guaranteed. CLC supplies the deterministic decision function and the stable reason-code registry those profiles can reference instead of re-inventing. Verifiable attenuation. A delegation chain must only narrow. Declaring attenuation is easy; verifying it at the receiving endpoint requires a shared relation. Contains (Section 13.4) plus its reason codes (child_exceeds_parent et al., Section 13.5) turn "only narrows" from an application-layer assumption into a checkable language-level fact. One model for authorization and evidence. The core deliberately uses the same identifier and constraint model on the authorization side (the grant) and, prospectively, on the evidence side (the observed action), so an audit trail can be reasoned about with the same relation that authorized it.

How CLC relates to adjacent work The comparison below is by role, not by feature count: the other works are carriers, profiles, or architecture, whereas CLC is the semantic layer. Two principles keep the relationship complementary rather than competitive: CLC defines evaluation, not carriage. CLC does not define token formats, trust models, or policy languages. The mapping from a carrier's native authorization structure to a CLC grant is the carrier profile's responsibility; CLC only fixes what the mapped grant then means. The core already ships consumption mappings. Appendix A ("Consumption Mapping") lists example consumers (AIC-JWT DA, RAR authorization_details, EMILIA AEB, delegation chains), and capability/data/_vectors/clc-v1/crosswalk-vectors.json pins executable cross-walk cases for the oauth-rar->clc-v1, aic-jwt-da->clc-v1, emilia-aeg->clc-v1, ucan->clc-v1, and delegation-chain->clc-v1 profiles. A carrier adopting CLC-D adds a containment case to that crosswalk rather than a new mapping mechanism. CLC-D ships its own such corpus: capability/data/_vectors/clc-d/containment-crosswalk-vectors.json maps a native representation (AIC-JWT DA, OAuth RAR, UCAN, delegation chain, and the adjacent drafts ATN, AAT, AIP, AAE, AOA, AEGIS) to a grant on each side and asserts Contains — the attenuation analogue of the CLC-A cross-walk. The mappings that are already pinned, and the ones a carrier would add, are: Carrier / native form Maps to CLC Relation Status AIC-JWT DA capability[] ({id, params, constraints}) one Grant per entry Entailment; CLC-D Contains for the C_agent ⊆ P_grants boundary pinned (aic-jwt-da->clc-v1, ccx-001..005, ccx-013) OAuth RAR authorization_details ({type, actions[], params}) one Grant per action (type:action) Entailment pinned (oauth-rar->clc-v1, ccx-006..007) UCAN {with, can} Grant {with:can} Entailment pinned (ucan->clc-v1, ccx-008..009) EMILIA AEG capability_class Grant {capability_class} Entailment pinned (emilia-aeg->clc-v1) Delegation chain (per-hop granted sets) Intersect of the links Intersection; CLC-D Contains per hop for attenuation pinned (delegation-chain->clc-v1, ccx-010..012) ATN Capability Manifest (draft-somoza-dmsc-atn-agent-trust-negotiation Section 5.1/Section 9.1/Section 9.2) one Grant per capability: atn/manifest-v1:<id>:<action>[:<schema_digest>]; resource_bounds and numeric conditions → params CLC-D Contains is the per-capability form of ATN's intersection; ATN's identity rule (same id and same schema digest) is carried by the digest path segment; preconditions are a union axis (Section 9.2), never compared pinned (atn-manifest->clc-v1, ccx-014..017, ccx-043..044) Attenuating Agent Tokens I4 (draft-niyikiza-oauth-attenuating-agent-tokens Section 4.5) one Grant per tool: aat/toolset-v1:<tool>; the argument-constraint map → params CLC-D Contains for tools(derived) ⊆ tools(parent) and the per-type constraint subsumption pinned (aat-i4->clc-v1, ccx-018..023) AIP attenuation walk (draft-prakash-aip Section 4.4) Grant aip/scope-v1:<capability>; budget ceiling → params CLC-D Contains for the per-dimension narrower-or-equal walk (scope, budget) pinned (aip-attenuation->clc-v1, ccx-024..029) AAE mandates and CONSTRAINTS (draft-kroehl-agentic-trust-aae Section 2.3/Section 3) Grant aae/mandate-v1:<action>; unwrapped CONSTRAINTS values → params CLC-D Contains for AAE's per-element "equal to or more restrictive" pinned (aae-constraint->clc-v1, ccx-030..034) AOA operation scope (draft-liu-agent-operation-authorization Section 6.2) Grant aoa/scope-v1:<operation> CLC-D Contains for the "scope string containment" the AS validates pinned (aoa-scope->clc-v1, ccx-035..037) AEGIS AIAM-1 delegation (aegis-initiative/aegis-governance, AIAM1-DEL-010) Grant aegis/action-v1:<domain>:<operation>; numeric context bounds → params CLC-D Contains for monotonic authority narrowing; allowed_roles/environment/risk_level/scope have no CLC field and are left to the carrier, not compared (see Appendix C) pinned (aegis-delegation->clc-v1, ccx-038..042) W3C ZCAPs capability document Grant (delegation schema/type → id; caveats → params/constraints) Entailment; CLC-D Contains for chained delegation profile to be written by the ZCAPs adopter Two mapping notes the pinned profiles make explicit, because they are places a reader could mistake a carrier's rule for the relation itself: A carrier's own "contains"/"subset" may name a constraint type, not the capability relation. AAT Section 4.5 defines argument-constraint types named contains and subset (a required-set superset and an allowed-set subset). Those are values inside params; the capability-level relation is still Contains, and the constraint axes compose by union across a chain. The same applies to AAE's allowed_domains and AEGIS's context bounds. A profile resolves a carrier's inheritance and defaults before mapping, and must not invent semantics the carrier does not state. AIP resolves an absent dimension to its nearest ancestor (Section 4.4), so the profile maps an absent dimension to no bound, not to a bound. AEGIS's dotted ids are hierarchical, but AEGIS grants capabilities individually and does not define domain-level containment, so the profile does not widen a bare domain into a namespace wildcard (ccx-042 fails closed). Contains is defined over the declared sets the profile emits: the profile carries the carrier's resolution, and what the carrier leaves undefined the profile leaves undefined — it does not fill the gap with an allow. The full contract is Section 13.8.1. The rows marked profile to be written are placeholders: this document does not guess another draft's field grammar. A carrier profile lands as a MapProfile case plus cross-walk vectors, exactly as the pinned profiles did. Work Role Relationship to CLC OAuth Rich Authorization Requests (RAR) Structured authorization_details carried in the authorization request RAR defines the carriage of structured authorization; CLC is a candidate evaluation language for the capabilities RAR declares. Attenuating-agent-token work observes that RAR expresses a request but does not itself define how a holder derives or verifies a narrower token — the layer Contains addresses. goes one step further and lets each domain schema define its own subset semantics; the carrier profile's job is to carry that decision into the mapped grant. supplies a named constraint vocabulary for such a profile (max_amount with amount and currency, recipient_domain), with no conversion between currencies OpenID Connect agent-identity claims Agent capability claims in an ID Token A profile could define how OIDC-delivered capabilities are evaluated and how a delegation chain is containment-checked WIMSE AI Identity Management System (, draft-ietf-wimse-aims) Informational best-practice framework reusing WIMSE and OAuth; explicitly identifies gaps rather than defining a capability algebra CLC is a candidate concrete evaluation language for the authorization step AIMS describes and a candidate answer to the "capability containment" gap it leaves open; the two are complementary, not competing WIMSE agent delegation chain (, draft-asor-wimse-agent-delegation-chain) Token format, chain linkage (par_hash), a scope/constraint vocabulary with subsumption rules, and an 8-step offline verification algorithm carried in an RFC 9068 JWT profile A carrier-level peer that defines how a narrower token is carried and verified; CLC is the carrier-neutral semantics for the per-hop subset judgement such a verifier performs. Contains is the relation its DT(child) ⊆ DT(parent) check needs, and its scopes/constraints map onto CLC's Grant/params/constraints; the two are complementary, not competing (the AAT row makes the same split one layer up) W3C ZCAPs Linked-Data-Proof signed capability documents with caveats and chaining ZCAPs defines the document/carrier; CLC can define the containment relation over its capabilities UCAN DID/IPLD authorization tokens with delegation and attenuation Same split: UCAN is the carrier, CLC the semantics AEGIS capability registry and AIAM-1 delegation (, aegis-initiative/aegis-governance) Hierarchical dotted capability registry, per-grant scope/constraints, a deterministic decision algorithm with verdicts (allow/constrain/escalate/deny), and monotonic authority narrowing (AIAM1-DEL-010); composition is explicitly not closed under transitivity (AIAM1-CAP-011) Closest in goal (capability declaration + deterministic evaluation + narrowing); differs in form — AEGIS is a governance architecture with a policy/registry layer, CLC a carrier-neutral decision function over grants. Contains is the relation AEGIS's monotonic-narrowing check needs; AEGIS's non-transitive composition rule is compatible (CLC-D Contains is also non-transitive: it is a per-pair predicate, not a closure) Agent Identity Protocol (, draft-prakash-aip) Delegation-chain token with a Datalog policy layer and a structural attenuation walk (V4) over scope, budget, time, domains, principal Overlaps CLC's entailment/intersection functionally, but pins a Datalog policy language. AIP Section 4.4 makes the same distinction CLC-D does — attenuation is a property of capability content, not of the append-only container — so CLC can be the shared deterministic semantics such a checker is validated against Agent Trust Negotiation (, draft-somoza-dmsc-atn-agent-trust-negotiation) Capability Manifest JSON with schema binding, dimension semantics, and a Capability Intersection Algebra (Section 9) with per-dimension rules including preconditions union Overlaps the capability-container target and defines an intersection; CLC-D supplies the single-pair containment predicate that runs before and alongside that intersection (Section 13.8). ATN's ordered dimension lattices (effects, external_calls, …) are the carrier-level analogue of CLC-D's mode lattice (Section 13.4.5), which stays out of the language relation Agent Operation Authorization (, draft-liu-agent-operation-authorization) Operation-proposal/authorization JWTs with a delegation_chain; the AS validates that a sub-operation is "strictly narrower in scope" (Section 6.2) via policy templates, OPA, or scope-string containment Same "no escalation beyond the original scope" goal; AOA's scope-string containment is exactly a carrier instance of Contains, and AOA's delegation_chain is the carrier for the chain CLC-D reasons over Attenuating Agent Tokens (, draft-niyikiza-oauth-attenuating-agent-tokens) Token chain with a capability lattice (C(child) ⊆ C(parent), Section 4.1) and six attenuation invariants; I4 defines per-type constraint subsumption with Decidable/Sound/Deterministic requirements (Section 3.5.1) The closest formal neighbour at the language level: C(child) ⊆ C(parent) is the property Contains decides, and AAT's naming of a constraint type contains is a caution that the relation and a constraint value must not be conflated Agent Authorization Envelope (, draft-kroehl-agentic-trust-aae) Verifiable Credential envelope with MANDATE/CONSTRAINTS/VALIDITY blocks and an explicit "equal to or more restrictive" definition per element (Section 3); warns of delegation amplification (Section 7.4) AAE defines the carrier blocks and a closed, deterministic constraint language; CLC-D's params_not_narrower / child_exceeds_parent are the stable reason codes that make AAE's "strictly subordinate" check reportable External Verifier Contract (, draft-kondoju-evc) Standardizes how an external verifier is invoked and returns a verdict (allow/deny/denial codes, fail-closed exit semantics) Orthogonal and complementary: EVC is the verdict interface, CLC-D is the decision semantics and its reason codes. An EVC implementation can compute CLC-D's verdict and surface the same reason codes Agent-auth architecture drafts (e.g. draft-klrc-aiagent-auth) "Agent as workload", reusing existing mechanisms; notes that no single existing policy engine covers the full delegation-chain verification need A natural consumer: CLC can be the evaluation language such a framework calls into Dual-identity / attenuating-token drafts (e.g. draft-ni-wimse-ai-agent-identity, AAT above) Bind agent identity to owner identity; define how a holder derives and a verifier checks a narrower token Answers who delegated and how derivation is carried; CLC answers what was delegated and whether it narrowed Agent interaction/delegation protocols (e.g. AIDP) Interaction and delegation flow over capability systems Capability-based sibling; CLC is the evaluation layer rather than the interaction flow WIMSE agent credential delegation (, draft-sweeney-wimse-credential-delegation) Server-side counterpart of a delegation-chain token: the same scope/constraint vocabulary applied where the credential is issued A carrier-level consumer like , one side of the exchange: the issuance side maps to grants, and Contains is the relation its attenuation check needs Cross-organisation delegation requirements (, draft-reece-wimse-cross-org-delegation) Requirements only — recursive attenuation (R1), a verdict vocabulary, and a separation of admission from consumption; no token format There is nothing to map, so it is cited as requirements, not as a profile: Contains per hop plus AuthorizeWithChain is a concrete answer to R1, and its verdict vocabulary is the state/reporting layer Section 13.9.7 addresses The external names above are recorded as context and MUST be re-verified against their current revisions before any submission or citation; this subsection makes no claim about their exact present contents. The mapping rows were checked against these revisions on 2026-09-21: draft-somoza-dmsc-atn-agent-trust-negotiation-00 (Section 5, Section 6, Section 9), draft-niyikiza-oauth-attenuating-agent-tokens-01 (Section 3.5, Section 4), draft-prakash-aip-02 (Section 3.3, Section 4.4), draft-kroehl-agentic-trust-aae-02 (Section 2.3, Section 2.5, Section 3), draft-liu-agent-operation-authorization-02 (Section 3, Section 6), draft-ietf-wimse-aims-00, and the aegis-initiative/aegis-governance repository (AIAM-1 v0.1, AIAM1-DEL-010), and draft-asor-wimse-agent-delegation-chain-02 (Section 3, Section 4, Section 5, Section 6, re-checked 2026-10-17 with AIP-02). Each of these is an individual draft, an informational draft, or a non-IETF repository, with one exception: the WIMSE working group adopted draft-ietf-wimse-aims on 2026-09-09, so it is cited here as a working-group document rather than as an individual draft. The AEGIS material is an informational architecture, not a specification with a formal standing.

Where the demand actually is The need for a shared evaluation layer is increasingly explicit in the authorization community: the recurring complaint is not a lack of carriage formats but a lack of agreement on what a carried authorization means and on how a delegation chain is verified to only narrow. The following distinction determines what "demand" should be taken to mean here: If demand means "adopted as a WG standard": uncertain. Several in-flight efforts (AIP, ATN, AOA, AEGIS, above) are attacking the same problem from different angles, and there is no consensus that a separate authorization language is wanted. A standalone individual draft is unlikely to be adopted directly on that basis alone. If demand means "used by implementers": the need is real and immediate. Agent frameworks and enterprise security teams each re-implement an authorization check and each ask the same question — "what may this agent actually do, and did the delegation only narrow?". A small, carrier-neutral, tested decision function plus corpus is directly reusable there. The strategic consequence is that adoption is earned by use, not awaited from a standards vote: the corpus and the reference implementations are the contribution, and the standards reference follows if and when downstream implementers cite it.

Adoption risks Carrier dependence. CLC only matters once a carrier (AIC, WIMSE, ZCAPs, UCAN, a RAR profile) binds it. The Section 13.8 AIC-JWT cross-walk is one such binding; without bindings CLC has no reach. Competing semantics, not a blank field. AIP, ATN, AOA, and AEGIS each define (or assume) evaluation semantics of their own. CLC enters a field with several incumbents, so it must be smaller (a decision function, not a policy language or registry), carrier-neutral, and tested to be worth citing. Ecosystem competition. A working group could prefer to define its own authorization meta-syntax rather than reference an external language. The response is scope discipline: CLC defines the evaluation and nothing else, and stays small enough to be cited. Implementation independence. The three current implementations share an author, so they are a regression test, not independent validation (Section 12, restated for CLC-D in Section 13.7). Independent implementation remains the gating risk for any standards claim.

Positioning statement CLC is best positioned not as a standalone standard but as the language specification other standards reference when they need to define authorization evaluation and delegation narrowing. Concretely: when an AIP-style Datalog checker, an ATN-style condition evaluator, or an AEGIS-style policy evaluator needs a shared, deterministic semantic to validate against or interoperate with, CLC is a candidate. Publishing CLC as the capability language core of a wider agent-authorization architecture (the AIC direction) is exactly that positioning; containment is kept additive so such a reference can be made without disturbing the CLC-A core any adopter already implements.

Carrier decision states that are not containment results This subsection is informative and adds no requirement. Review of the carrier mappings surfaced a recurring modelling question: adjacent work introduces decision states that are neither a containment verdict nor an authorization verdict, and a carrier that adopts CLC should not be pushed into inventing a fourth verdict for them. The table records where each such state lands in CLC's own vocabulary. Nothing here changes Section 9, Section 8.4, Section 8.5 or Section 13.4. A carrier's own state Where it is defined (recorded 2026-10-16) CLC reading refusal is permanent for the presented artifact Section 5 deny with its own Section 9.2 reason if the artifact fails evaluation; the durability of the refusal is the verifier's state, not a CLC value refusal is transient because an input could not be obtained Section 5 a core decision cannot reach it: an input the core does not hold is a deny under Section 6 (fail-closed), while a recognized obligation the core does not evaluate is allow_unresolved plus a Section 8.4 residual evaluation.status = not-evaluated with unavailableInput Section 5.4 allow_unresolved with the residual obligation named; never a fourth verdict. evaluation.status = evaluated with decision.reported = deny is deny per-input established-negative / not-established / stale Section 6 violated / unknown / unknown respectively at the Section 8.5 Resolve boundary; stale for a non-time input has no core clock, so it is unknown until the carrier refreshes it undecided membership or attenuation (UNDECIDED) Section 6 allow_unresolved for the undecided part, with the remainder decided normally (Section 8.4, Section 8.5) local-only / not decidable without a source the decision point does not hold Section 6 allow_unresolved residualized to the consumer; the core is locally decidable by design (P9) and never performs the lookup Two consequences a carrier should take from the table. First, the three CLC verdicts are closed: a carrier maps its states onto allow / deny / allow_unresolved and reports its own subdivision alongside, in its own record format — the evaluation member of is the worked example. Second, a state that requires a source the decision point does not hold is a decision to defer, not a new verdict: the Section 8.4 residual channel exists precisely so that "I could not check this yet" travels with the decision instead of being flattened into allow (which would grant more than the artifact declares) or into deny (which would misreport a policy decision).

Open Issues Section 13.10 records the consciously-deferred gaps surfaced in review; the items already landed as core revisions are marked closed, the rest are recorded for later discussion. Items marked candidate v1.x are candidates for the core language to adopt without breaking CLC-A inputs; items under "CLC-D" would extend containment itself.

Parameter model rigidity (closed in CLC-1.10) Landed as Section 6.5, the optional param_bounds field: Number: inclusive min/max intervals and a step multiple rule. Enums: array membership plus min_items/max_items cardinality. Optional keys: an optional marker exempts a declared key from the omission half of layer 7. Defaults: a param_defaults grammar with the precedence explicit > default > absent. The rejected shape is a {min,max} object inside params — it would collide with object recursion (Section 6.2), so the bounds live in a sibling field and no existing grant changes meaning.

Consumer obligations for allow_unresolved (closed in CLC-1.11) Section 8.4 delivers residual obligations on an allow_unresolved verdict; Section 8.5 now defines the consumer's feedback loop and this item is closed: Resolve(decision, resolutions, now?) -> Decision, with each unresolved constraint reported as satisfied / violated / unknown; a propagation rule for partially-evaluated sets (all satisfied → allow, any violated → deny with {type}:violated, remainder → allow_unresolved); staleness/TTL semantics for time:window residuals: with now, the core clock evaluates the window and the discharge horizon is the current segment's end, so a cached allow expires with the window. This went beyond the original "Resolve is core, TTL is profile policy" split: both are now core (Section 8.5). The coarser identity-level consumer gate Discharge (the reference implementations' helper) remains as the satisfied-only case.

Constraints as a union axis (closed in CLC-1.12) This revision makes an explicit design decision: constraints are not part of the containment relation (Section 13.4.4). They compose by union (conjunction) along a chain, which Intersect already implements, and no Contains layer reads them. Contains therefore stays the smaller relation — containment over (identifier, parameters) — and the "what does the chain collectively require?" question is answered by the separate derived function ConstraintUnion (Section 7.1, new in CLC-1.12). Folding the union into Contains was considered and rejected: it would make the relation no longer a pure subset on the declared tuple, and a null constraint check would then be able to hide a broken identifier/parameter boundary. A consumer wanting "the child's whole authority is inside the parent's" composes the two: Contains per hop plus ConstraintUnion over the chain.

Containment as evidence, not authorization (closed in CLC-1.13) Section 13.4 keeps containment a declared-set comparison. A delegation certificate binds a child grant; an operation-time authorization still needs the decision function (Section 9). CLC-1.13 adds the fourth relation AuthorizeWithChain(chain, op) (Section 13.11), the one-call chain check named here: it evaluates Contains per hop and then Authorize against the intersection of the chain, so the parent's constraints (a union axis, outside containment) are not lost. It is a CLC-D function, not a core change; a CLC-A implementation is unaffected.

Profile obligations for a chain claim (closed in CLC-1.18) Cross-checking the eleven pinned carrier mappings raised two questions the profile contract did not answer; both are now stated in Section 13.8.1 instead of being left to each profile's prose: On what basis may a profile claim transitivity? Contains is per-pair (Section 13.3), and the Section 13.8.1 non-goal list already said a chain check is the profile's iteration — but it did not say what a profile owes a reader when it performs that iteration. Section 13.8.1 obligation 6 now requires the basis to be declared (definition-derived, mechanically-checkable or asserted, with the finite domain and deterministic procedure stated for the second) and forbids reading two adjacent Contains results as a transitive one. What happens to a carrier's conjunctive fields? Section 13.4.4 already keeps constraints outside the relation and Section 7.1 already exposes the union projection, but a profile could still describe a chain as contained while carrying a hop's preconditions (draft-somoza-dmsc-atn-agent-trust-negotiation Section 9.2), CONSTRAINTS (draft-kroehl-agentic-trust-aae Section 3) or requires (draft-izmaylov-agent-permission-receipts Section 5) uncompared. Section 13.8.1 obligation 7 now forbids reporting a union axis as a containment result. Where do a carrier's own decision states go? The new informative Section 13.9.7 records the mapping onto the three verdicts and the Resolve values so a carrier is not pushed into inventing a fourth verdict. All three bind profiles, not the relation: no CLC-A or CLC-D verdict, reason code or vector moves, and an implementation that publishes no profile is unaffected. What changes is what a profile may claim: one written against the earlier text either adds the declarations or states that it claims no chain check.

AuthorizeWithChain (fused chain authorization) Contains is a declared-set comparison and Authorize is an operation-time decision; a delegating consumer that has a chain often wants both in one call. AuthorizeWithChain is that convenience, defined at the CLC-D layer:

AuthorizeWithChain(chain, op) → Decision // chain = ordered Grant[], root first

An empty chain fails closed: deny("absent_source") (Section 7 rule 5). For each adjacent pair (chain[i], chain[i+1]), evaluate Contains (Section 13.4). The first hop that is not contained ends the call with deny(reason), where reason is that hop's Section 13.5 code — child_exceeds_parent or params_not_narrower, the two codes the relation can return; delegation_mode_not_narrower never appears here, because the chain gate calls Contains, which takes no mode argument (Section 13.4.5). This chain gate runs before op validation: a broken chain is reported even when the operation is also absent, because the chain is the subject of this function. Otherwise compute the effective chain grant G = Intersect(chain...) (Section 7). Because constraints are outside containment (Section 13.4.4), this step is what brings every ancestor's params and constraints into force — authorizing against the leaf grant alone would let an operation pass that violates an ancestor's constraints (the union axis is not in Contains). An Intersect refusal (no_overlap, empty_bound_denies_class, invalid_params_binding) is returned as deny(reason). Return Authorize(G, op) (Section 9) unchanged — allow / allow_unresolved / deny with its own Section 9 reason codes. Caller obligation. AuthorizeWithChain evaluates the chain as presented: the caller MUST supply the complete, authenticated, root-first chain. The function fetches no missing link, verifies no signature or trust anchor, and detects no truncation or reordering — a verdict over a truncated, reordered or unauthenticated chain is a verdict about the presented sequence, not about the delegation it does not carry (authentication is the carrier's concern, Section 11). This obligation adds no decision rule: the steps above are unchanged by it. AuthorizeWithChain is a CLC-D function: it is not part of CLC-A, and an implementation claiming only CLC-A is unaffected. It introduces no new core semantics — it fixes the order of two existing relations and refuses fail-closed at each step. A chain carrying param_bounds is authorizable: step 3's Intersect(chain...) combines the hops' bounds with the Section 6.6 meet, so an ancestor's bound (e.g. max:100) stays in force over a narrower child (e.g. max:50). Because Section 13.4.3 requires the declaration site of every key to match at each hop, a valid chain never presents the cross-site case Section 6.6 refuses. The two-grant form AuthorizeWithChain(parent, child, op) named in Section 13.10.4 is the degenerate case chain = [parent, child].

Revision and Governance Containment is folded into this document's revision stream: its changes are recorded in the Revision History and its conformance class CLC-D is declared in Section 12.1 in step with the language revision (this revision is CLC-1.19; CLC-D first appeared in CLC-1.9, folded from EXT-00 rev 0). A CLC-A input is unaffected by the addition of CLC-D; compatible reading of CLC-A inputs is the floor. CLC-D adoption is per-implementation: an implementation may claim CLC-A without claiming CLC-D. The corpus (64 containment vectors, 784 property cases, 45 cross-walk vectors, 15 AuthorizeWithChain vectors) is a draft snapshot; the README in capability/data/_vectors/clc-d/ maintains the live count and the date the snapshot was generated.

Security Considerations Fail-closed: undefined/malformed/unknown → deny. Deny-when-declared: empty bounds deny the class. No canonical broadening: segment-boundary, not lexical prefix. Composition narrows only: an intersection removes authority. Whether a delegated grant stays inside its parent's boundary is the containment question Section 13 answers (Contains); intersection alone (Section 7) does not answer it, and a delegation profile MUST NOT substitute one for the other. Containment is fail-closed by construction: every Contains layer defaults to false; a grant pair any layer cannot validate is refused, with no warning-then-allow path (Section 13.4). Containment is not a constraint oracle: Contains does not read constraints (Section 13.4.4). A consumer that uses Contains as its only gate MUST compose the delegation chain's Intersect so the parent's constraints stay in force; otherwise a child that omits a parent constraint could be admitted while the constraint is unenforced. allow_unresolved is orthogonal to containment and MUST NOT be read as a containment result. Containment reason-code leakage: child_exceeds_parent reveals that a child requested an identifier the parent does not cover. A boundary that must not leak policy shape MAY collapse that single code into capability_not_authorized (never params_not_narrower, Section 13.5). Mode lattice must be carrier-pinned: an implementer that maps authorized/representative the wrong way round inverts the boundary; Section 13.8 pins the AIC-JWT ordering, and other carriers MUST pin theirs in the profile that adopts CLC-D. Stable reason codes: same input → same reason across implementations. Evidence binding is separate from native verification: Match checks content correlation; native verification is the consumer's responsibility. Parsing divergence must not change the decision: params are normalized at the input boundary per Section 6.2 (JCS serialization, duplicate keys, non-finite/over-precision numbers, size/depth caps). A consumer that decodes into a re-orderable map and re-encodes loses duplicate keys and cannot represent non-finite numbers; two such consumers would reach different verdicts on the same raw input. Decisions are made on the boundary-validated form, not on a lossy re-serialization. Recognized-but-unevaluated is not silent acceptance: a constraint the core recognizes but cannot evaluate MUST appear in the decision's unresolved field — never dropped (Section 8.4). The consumer must evaluate or confirm each such constraint before acting, otherwise it MUST deny (AAC Section 6.6). A core-clock discharge is time-bounded: when Resolve (Section 8.5) evaluates a time:window obligation with a supplied now, the discharge is valid only for the segment containing now. A consumer MUST NOT cache the resulting allow past that segment's end — it re-invokes Resolve with a current now before acting, or derives the horizon from the segment end. Caching a clock-based allow turns a window into an unbounded permit. Reason-code detail suffix is diagnostic-only: everything after the first : (e.g. the offending param name) MUST NOT change the verdict and MUST NOT be relied upon for decisions. Consumers match on the code prefix before the : (Section 9.2). Revision mismatch is fail-closed: an incompatible language revision (Section 12.1) yields deny("unsupported_language_revision") resolved before any layer — never a silent downgrade or best-effort re-interpretation. Resource exhaustion is bounded at the input boundary: the 512-byte serialized-size cap and the depth-32 nesting cap (Section 6.2 step 4) apply to raw_params as much as to every other input, keeping recursive evaluators safe from deep-nesting and oversized-params blowup.

IANA Considerations This document requests no IANA actions. Constraint types (max_rows, time, network) and reason codes are defined by this document as fixed sets. CLC-1.19 adds two reason codes to that set (unit_mismatch, unsupported_domain) and no new constraint type. Should this work be adopted by a working group, that group may wish to consider whether either set warrants a registry; this revision does not propose one. The containment relation (Section 13) registers three additional reason codes (child_exceeds_parent, params_not_narrower, delegation_mode_not_narrower — the last produced by the binding profile's delegation-mode pre-check, Section 13.4.5, never by the relation itself); they are part of the same fixed set, and no constraint reason code is defined for containment because constraints are outside the relation (Section 13.4.4).

Privacy Considerations The language itself transports and stores nothing. Privacy exposure comes from what carriers put into it and from what evaluators report: Capability identifiers and parameter values describe policy. They can reveal organizational structure, service topology, network ranges (network constraints), working hours (time windows), tenant names, or purposes. Deployments should treat grants as policy-confidential material. Distinct reason codes reveal the shape of a grant: the difference between params_missing, undeclared_param and not_in_enum tells an observer what the grant constrains. Where the requester is untrusted, a consumer should consider collapsing reason codes at the boundary, as this specification already does for identifier-level failures (Section 9.1). Parameter values may carry personal data if a scheme defines them that way. Scheme authors should avoid personal identifiers as parameter names or values. Residual obligations (unresolved, Section 8.4) and any audit record built from decisions can persist policy and usage information; retention is the carrier's responsibility (Section 11). The reference corpus published with this document is synthetic and contains no personal data. Containment reasons (Section 13) are emitted to the child's requestor at the delegation step, not to end-users, and the child receives the failure code, not the parent's declared bounds. Where bounds themselves are sensitive (e.g. network/scope declarations), a profile SHOULD log codes, not values.

In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. This document defines a date and time format for use in Internet protocols that is a profile of the ISO 8601 standard for representation of dates and times using the Gregorian calendar. I-JSON (short for "Internet JSON") is a restricted profile of JSON designed to maximize interoperability and increase confidence that software can process it successfully with predictable results. JavaScript Object Notation (JSON) is a lightweight, text-based, language-independent data interchange format. It was derived from the ECMAScript Programming Language Standard. JSON defines a small set of formatting rules for the portable representation of structured data. This document removes inconsistencies with other specifications of JSON, repairs specification errors, and offers experience-based interoperability guidance. Cryptographic operations like hashing and signing need the data to be expressed in an invariant format so that the operations are reliably repeatable. One way to address this is to create a canonical representation of the data. Canonicalization also permits data to be exchanged in its original form on the "wire" while cryptographic operations performed on the canonicalized counterpart of the data in the producer and consumer endpoints generate consistent results. This document describes the JSON Canonicalization Scheme (JCS). This specification defines how to create a canonical representation of JSON data by building on the strict serialization methods for JSON primitives defined by ECMAScript, constraining JSON data to the Internet JSON (I-JSON) subset, and by using deterministic property sorting. This document specifies a new parameter authorization_details that is used to carry fine-grained authorization data in OAuth messages.

Consumption Mapping The rows below are examples of consumers of this shared vocabulary, not required profiles: conformance to CLC-A does not depend on any of them. RAR authorization details are one such carrier; CLC is a candidate evaluation language for the capabilities they declare. Consumer Grammar Binding Verdict Notes AIC-JWT DA capability[].id Entailment (Section 6.1) Decision (Section 9) AIC-JWT Section 5 binding EMILIA AEB AEG capability_class Match (Section 6.4) + Entailment (Section 6.1) SATISFIED (Section 10) + Decision (Section 9) AEB Section 3 decision levels (VERIFIED/MATCH/SATISFIED) + Section 5.1 ObservedAction + Section 7 AEC slots; a VERIFIED authorization artifact carries the Grant RAR authorization_details type="capability" Entailment (Section 6.1) Decision (Section 9) format Delegation chain each hop's declared set Intersection (Section 7) Decision (Section 9) intersection over declared sets only; a hop that must stay inside its parent is checked with Containment (Section 13) Delegation containment parent and child boundaries Contains (Section 13) Containment verdict (Section 13) per-hop child ⊆ parent; carrier vocabularies and reason-code mapping in Appendix C

Reference Vectors Grouping vs kind mapping: The appendix groups vectors by semantic category (B.1–B.6). The machine-readable vectors.json uses a kind field that collates these groups differently: kind=entail (65) covers B.2 (8), the 53 params vectors that sit under kind=entail (B.3's 70 rows minus params-028/-029/-034/-036 and the thirteen type-sensitivity backfill rows params-040–params-052, all seventeen being kind=decide), and the four scheme stress-test entail vectors (clinical-001/-002, payments-001, data-002); kind=decide (63) covers the B.5 rows below (34), the seven combined decision vectors, payments-002 and data-001, the four params boundary decisions that sit under kind=decide (params-028/-029/-034/-036, all four also listed in B.3), the three nested key-closure vectors (nested-001/-002/-003) and the thirteen type-sensitivity backfill vectors (params-040–params-052, also listed in B.3); kind=intersect (17) covers B.4 (10), the four combined vectors that call the intersect function (combined-004/-005/-008/-011) and the three CLC-1.15 cross-type audit vectors (intersect-011/-012/-013, outside the B.1–B.6 tables); kind=syntax (9) is exactly B.1. These counts are reproducible from the corpus itself: every vector in vectors.json carries its kind, so the mapping is machine-checkable rather than maintained by hand.

Syntax (9 vectors) # Input Expected Derivation S1 std/database-v1:query:SELECT valid literal identifier S2 std/database-v1:query:* valid trailing wildcard S3 *:query:SELECT deny bare * illegal S4 std/database-v1:query:SEL* deny partial segment wildcard S5 std/database-v1:query:{read,write} deny alternation not v1 S6 std/database-v1:query:[a-z] deny character class not v1 S7 database:query deny(invalid_capability_id) Section 3 scheme grammar: no vendor "/" product "-v" major (the spec's own counter-example) S8 bad:op deny(invalid_capability_id) Section 3 scheme grammar: scheme bad does not match vendor/product-vN (snips the lax-intake hole) S9 std/data-v1:fetch:item:42 valid multi-segment action + conforming scheme (positive boundary)

Entailment (8 vectors) # Grant Operation Expected Derivation E1 std/database-v1:query:SELECT std/database-v1:query:SELECT allow literal match E2 std/database-v1:query:* std/database-v1:query:SELECT allow trailing wildcard E3 std/database-v1:query:* std/database-v1:query:SELECT:deep allow wildcard multi-segment E4 std/database-v1:query:* std/database-v1:admin:DDL deny different namespace E5 std/database-v1:query:* std/database-v1:query deny no trailing segment E6 std/database-v1:query:SELECT std/database-v1:query:INSERT deny literal mismatch E7 std/database-v1:* std/database-v1:query:SELECT deny class-position (product-segment) wildcard is v1-forbidden: only a trailing action segment may be * (Section 3; Section 9.1 layer 3) E8 std/database-v1:* std/database-v1:admin:DDL deny same class-position wildcard; the v1-forbidden shape denies regardless of the action it faces (Section 3; Section 9.1 layer 3)

Params (70 vectors) The params vectors are grouped by the property each subgroup pins.

Basic bounds, enums, objects and unconstrained grants (P1–P15) # Grant Operation Expected Derivation P1 {"limit":100} {"limit":50} allow 50 ≤ 100 P2 {"limit":100} {"limit":150} deny 150 > 100 P3 {"tables":["a","b"]} {"tables":["a"]} allow subset P4 {"tables":["a"]} {"tables":["a","b"]} deny "b" absent (not_in_enum) P5 {"columns":{"t":["id"]}} {"columns":{"t":["id","name"]}} deny "name" absent (not_in_enum) P6 {} {"limit":50} allow unconstrained ({} ≡ absent; the literal {} is pinned by decide-028, params-006 pins the absent form) P7 {"tables":[]} {"tables":["a"]} deny explicit empty bound denies the class P8 {"limit":100} {"limit":null} deny null invalid in v1 (invalid_params_null) P9 {"station":[1,2,3]} {"station":2} allow scalar member of array set P10 {"station":[1,2,3]} {"station":9} deny 9 ∉ allowed set (not_in_enum) P11 {"station":[1,3]} {"station":[1,3]} allow array, every element a member P12 {"station":[1,3]} {"station":[1,2,3]} deny 2 ∉ allowed set (not_in_enum) P13 {"station":[1],"speed":0.3} {"station":1,"speed":0.25} allow member + number bound unaffected P14 {"station":[3]} {"station":2} deny categorical: granting 3 does not cover 1/2 (not_in_enum) P15 {"columns":{"t":["id","name"]}} {"columns":{"t":["id"]}} allow object recursion: request element is a member of the granted set (Section 6.2 v1.1)

Raw-input rejection and reason precedence (P16–P27) # Grant Operation Expected Derivation P16 {"limit":100} raw {"limit":100,"limit":150} deny(invalid_params_duplicate_key) duplicate JSON key rejected at input normalization (Section 6.2 step 2) P17 {"limit":100} raw {"limit":1e400} deny(invalid_params_number) non-finite number literal (Section 6.2 step 3) P18 {"s":"x"} raw {"s":"<600 chars>"} deny(invalid_params_size) serialized form > 512 B (Section 6.2 step 4) P19 {"d":0} raw depth-33 object deny(invalid_params_size) nesting beyond depth 32 (Section 6.2 step 4) P20 {"s":"x"} raw {"s":"<512 B>"} allow serialized = exactly the 512 B limit (≤); positive boundary of P18 (Section 6.2 step 4) P21 {"d":0} raw depth-32 object allow depth exactly the 32 limit (≤); positive boundary of P19 (Section 6.2 step 4) P22 {"flag":true} {"flag":1} deny(params_exceed_grant) boolean is exact and is NOT a number: 1 must not satisfy a granted true (params-022) P23 {"flag":true} {"flag":true} allow positive side of P22 (params-023) P24 {"x":null} (no params) deny(invalid_params_null) layer 6 (null) precedes layer 7 (presence) (params-024) P25 {"s":"x"} raw {"s":"<512 B>","s":"dup"} deny(invalid_params_size) multi-fault: size (4) precedes duplicate keys (2) (params-025) P26 {"s":"x"} raw {"n":1e400,"s":"<512 B>"} deny(invalid_params_size) multi-fault: size (4) precedes number shape (3) (params-026) P27 {"a":1} raw {"a":1,"a":2,"n":1e400} deny(invalid_params_duplicate_key) multi-fault under the size limit: duplicate keys (2) precede number shape (3) (params-027)

Canonical size and Unicode boundaries (P28–P39) # Grant Operation Expected Derivation P28 {} raw {"n":1e-6,"s":"<494 a>"} deny(invalid_params_size) the received text is 511 octets but the JCS form writes 1e-6 as 0.000001, so the Section 6.2 step 4 size is 515 (params-033) P29 {} {"n":1e-6,"s":"<494 a>"} (decoded) deny(invalid_params_size) decoded counterpart of P28: both boundaries agree (params-034) P30 {} raw {"n":1.0,"s":"<498 a>"} allow the received text is 514 octets but the JCS form writes 1.0 as 1, so the size is 512 and inside the cap (params-035) P31 {} {"n":1.0,"s":"<498 a>"} (decoded) allow decoded counterpart of P30: a raw check counting the received token would refuse it (params-036) P32 {} raw {"s":"<100×U+1F600>"} allow 100 literal astral characters are 408 JCS octets; counting UTF-16 code units would double the count and refuse (params-037) P33 {} raw {"s":"<100×\ud83d\ude00>"} allow escaped spelling of P32: literal and escaped forms of one string MUST reach the same verdict and the same size (params-038) P34 {} raw {"s":"a\nb"} (literal U+000A) deny(invalid_params_number) a literal control character is not valid JSON text; Go and Python refused it already (params-039) P35 {} {"x":"<260×é>"} (decoded) deny(invalid_params_size) 260 U+00E9 code points serialize to 528 JCS octets > 512; non-ASCII sizes are measured on the canonical UTF-8 form, never in code points or UTF-16 units (params-028, rev CLC-1.4) P36 {} {"x":"<251×é>"} (decoded) allow 251 U+00E9 serialize to 510 octets ≤ 512; the positive side of P35 (params-029, rev CLC-1.4) P37 {"s":"x"} raw {"s":"<251×\u00e9 escapes>"} allow the same 510-octet string spelled with \u00e9 escapes reaches the same verdict and the same size as the literal form of P36 (params-030, rev CLC-1.4) P38 {"limit":100} raw {"s":"\ud800"} deny(invalid_params_number) a lone surrogate escape is not valid Unicode: refused at the raw boundary, never repaired to U+FFFD (RFC 8785 Section 3.2.2.2; Section 6.2 step 2; params-031, rev CLC-1.6) P39 {} raw {"s":"\ud83d\ude02"} allow a valid surrogate pair is one character (U+1F602, four UTF-8 octets) and counts as such under the size rule (params-032, rev CLC-1.6)

Type-sensitive equality (P40–P52) # Grant Operation Expected Derivation P40 {"n":50} {"n":"50"} deny(params_exceed_grant) a JSON string is not a number: "50" must not satisfy a granted 50 — Section 6.5 layer-8 type sensitivity inside the Section 6.2 algebra; host carriers normalize per Section 6.2 step 9 (params-040) P41 {"n":"50"} {"n":50} deny(params_exceed_grant) a number never satisfies a string grant — string row, exact; the reverse direction of P40 (params-041) P42 {"n":100} {"n":100.0} allow 100.0 is the same JSON number as 100 after step-1 canonicalization; 100 ≤ 100 (params-042) P43 {"n":100} {"n":99.5} allow a numeric grant is an upper bound, not an exact value: 99.5 ≤ 100 (params-043) P44 {"n":100} {"n":100.5} deny(params_exceed_grant) 100.5 > 100 (params-044) P45 {"n":100.5} {"n":100} allow the upper bound in the other direction: 100 ≤ 100.5 (params-045) P46 {"n":100.5} {"n":101} deny(params_exceed_grant) 101 > 100.5 (params-046) P47 {"s":"abc"} {"s":"abc"} allow string-branch guard: equal strings still allow under the type-sensitivity backfill (params-047) P48 {"flag":true} {"flag":"true"} deny(params_exceed_grant) boolean is exact only: the string "true" is not the boolean true (params-048) P49 {"n":1} {"n":true} deny(params_exceed_grant) true ≠ 1 (Section 6.5 layer 8) — the boolean-side reverse of P22 (params-049) P50 {"cfg":{"limit":"10"}} {"cfg":{"limit":10}} deny(params_exceed_grant) nested objects recurse with the same type sensitivity (mirrors nested-003) (params-050) P51 {"ids":[1,2]} {"ids":[1,"2"]} deny(not_in_enum) enum membership is type-sensitive: the member 2 is not the string "2" (params-051) P52 {"ids":[1,2]} {"ids":[1.0]} allow 1.0 is the member 1 after canonicalization (Section 6.5 layer 8) (params-052)

I-JSON bound and BOM (P53–P62) # Grant Operation Expected Derivation P53 {"n":9007199254740991} raw {"n":9007199254740991} allow top of the I-JSON integer range: 2^53 − 1 is the largest integer-valued magnitude accepted (params-053, rev CLC-1.16) P54 {"n":9007199254740992} raw {"n":9007199254740992} deny(invalid_params_number) the first refused integer-valued magnitude above the bound (params-054) P55 {"n":9007199254740993} raw {"n":9007199254740993} deny(invalid_params_number) 2^53 + 1 rounds to 2^53 in binary64; refused, never rounded (params-055) P56 {"n":9007199254740993} raw {"n":9.007199254740993e15} deny(invalid_params_number) the bound is on the mathematical value: the exponent spelling cannot launder an over-bound integer (params-056) P57 {"n":9007199254740993} raw {"n":9007199254740993.0} deny(invalid_params_number) a trailing .0 does not make it a fraction (params-057) P58 {"n":-9007199254740993} raw {"n":-9007199254740993} deny(invalid_params_number) the rule is on magnitude: the negative side is refused like the positive (params-058) P59 {"n":4503599627370495.5} raw {"n":4503599627370495.5} allow a non-integer near the top is unaffected: step 1 already fixed its rounding (params-059) P60 {"n":<30-digit>} raw deny(invalid_params_number) far over the bound and over-precision; both clauses resolve to one stable code (params-060) P61 {"n":1} raw with a leading UTF-8 BOM deny(invalid_params_number) a leading BOM is not part of a JSON text — refused before parsing (Section 6.2 step 8; params-061) P62 {} raw: BOM-prefixed and over the size cap deny(invalid_params_number) the BOM refusal outranks the step-4 size cap (params-062)

Numeric corner cases (P63–P70) # Grant Operation Expected Derivation P63 {"n":0} raw {"n":-0} allow -0 and 0 are one JSON number; JCS renders both as 0 (params-063, rev CLC-1.17) P64 {"n":1000} raw {"n":1e3} allow 1e3 ≡ 1000: number identity is the value, not the spelling (params-064) P65 {"n":0.0015} raw {"n":1.5e-3} allow 1.5e-3 ≡ 0.0015 (params-065) P66 {"n":1e100} raw {"n":1e100} deny(invalid_params_number) 1e100 is integer-valued and far outside the I-JSON range; binary64 cannot represent it exactly, but the mathematical-value bound refuses it independently of carrier representability (params-066) P67 {"n":1.7976931348623157e308} raw deny(invalid_params_number) the largest finite binary64 is integer-valued and over the bound → refused (params-067) P68 {"n":5e-324} {"n":0} allow a subnormal is finite and in range: 0 ≤ 5e-324 (params-068) P69 {"n":0} {"n":5e-324} deny(params_exceed_grant) the inverse bound direction: 5e-324 > 0 (params-069) P70 {"n":-0} {"n":0} (decoded) allow decoded -0 is the same number as 0 (params-070)

Intersection (10 vectors) Shorthand: params shown compact; constraints use colon notation. # Source A Source B Expected Derivation I1 {"tables":["a","b"]} {"tables":["a"]} {"tables":["a"]} overlap I2 {"tables":["a"]} {"tables":[]} deny deny-when-declared I3 (unconstrained) (no grant) deny absent source I4 {"limit":100} {"limit":50} {"limit":50} tighter bound I5 (unconstrained) constraint time:window:[{"start":"00:00","end":"01:00"}] recognized, not evaluated → unresolved carried on allow_unresolved verdict constraint added I6 {"tables":["a"]} {"tables":["b"]} deny no overlap I7 (no source / null) — deny(absent_source) zero sources: fail-closed (Section 7 rule 5) I8 {"limit":50} {} {"limit":50} empty params declares no constraint → bound preserved (bounded then empty, Section 7 rule 6) I9 {} {"limit":50} {"limit":50} source order must not matter (empty then bounded, Section 7 rule 6) I10 {}, id query:SELECT {}, id query:* {} narrower identifier wins; identifier comparison is params-free (Section 7 rule 2)

Decision (34 vectors) # Scenario Expected Derivation D1 valid grant, valid op, constraints pass allow all checks pass D2 no matching grant deny("capability_not_authorized") fail-closed D3 unknown constraint type deny("unknown_constraint") fail-closed D4 malformed capability_id deny("invalid_capability_id") input validation D5 same input twice same output deterministic D6 constraint violation deny("{type}:violated") constraint fail D7 grant bounds a param, operation omits it deny("params_missing") Section 6.3 step 4 fail-closed D8 operation param value is null deny("invalid_params_null") Section 6.2 null rule; may carry : <param> detail (Section 9.2) D9 bounded grant, operation has no params field at all deny("params_missing") Section 6.3 step 4 fail-closed D10 Authorize called with an absent/empty grant deny("capability_not_authorized") Section 9 fail-closed, no exception D11 input declares CLC-1.0 against a CLC-1.3 implementation allow same major, 1.0 ≤ 1.3 → compatible (Section 12.1) D12 input declares CLC-2.0 against a CLC-1.3 implementation deny("unsupported_language_revision") different major → fail-closed (Section 12.1) D13 operation carries a param key the grant does not declare (key closure) deny("undeclared_param") Section 6.2 key closure, Section 9.1 layer 7 request side D14 both a missing grant key and an undeclared request key deny("params_missing") layer-7 order: missing before undeclared D15 absent/empty grant and absent operation deny("capability_not_authorized") Section 9.1 pre-check resolves before any layer, incl. the absent-operation case D16 grant valid, operation has no id deny("missing_capability_id") layer 1 D17 grant carries time:window with a cross-midnight single segment (22:00→06:00) deny("invalid_constraint") a single segment crossing midnight is out of grammar (decide-019) — a crossing must be split into two segments D18 grant carries network:cidr (array form, IPv4/IPv6) allow_unresolved, unresolved:[<constraint>] recognized, no core evaluator → residual obligation (Section 8.4; decide-020) D19 time:window scalar second form (window:3600) deny("invalid_constraint") out of Section 8.1 value grammar (decide-021) D20 network:cidr without prefix length deny("invalid_constraint") out of Section 8.1 value grammar (decide-022) D21 max_rows constraint, op carries no max_rows value deny("max_rows:violated") fail-closed op-absent (decide-023) D22 time:window split-form multi-segment window (22:00→00:00 + 00:00→06:00) allow_unresolved, unresolved:[<constraint>] cross-midnight split-segment grammar (decide-024) D23 op scheme bad passes no vendor/product-vN deny("invalid_capability_id") Section 3 scheme grammar (decide-025) D24 grant id bad:op against a valid operation deny("capability_not_authorized") Section 3 fails in Entails; ID-level reason collapses (decide-026) D25 max_rows exactly at the bound allow violation is strict >; core-evaluated so no unresolved (decide-027) D26 grant params:{}, op any params → unconstrained allow {} ≡ absent (decide-028, Section 9.1) D27 multi-grant: G1 {limit:10} denies, G2 {limit:100} allows allow any-one-covers authorizes (decide-029, Section 9.1) D28 multi-grant: G1 {limit:10} + G2 {limit:6}, op {limit:50} deny("params_exceed_grant") all covering grants reject → first reason in canonical order (decide-030, Section 9.1) D29 multi-grant allow_unresolved: G1 network:cidr + G2 time:window allow_unresolved, unresolved:[both] residual obligations union across covering grants (decide-035, Section 9.1/Section 8.4) D30 operation id uses a forbidden wildcard shape (*:query:SELECT) deny("unsupported_wildcard") wildcard-shape detection precedes the base grammar, and layer 1 propagates the specific code rather than invalid_capability_id (decide-018, Section 3/Section 9.1) D31 grant bounds max_rows:10, operation carries max_rows:"garbage" deny("max_rows:violated") op-side value outside the Section 8.1 domain (finite non-negative integer) fails closed, never passes unchecked (decide-031, rev CLC-1.4) D32 grant bounds max_rows:10, operation carries max_rows:true deny("max_rows:violated") boolean is outside the Section 8.1 domain (decide-032, rev CLC-1.4) D33 grant bounds max_rows:10, operation carries max_rows:-1 deny("max_rows:violated") a negative is outside the Section 8.1 domain (decide-033, rev CLC-1.4) D34 grant bounds max_rows:10, operation carries max_rows:1.5 deny("max_rows:violated") a fraction is outside the Section 8.1 domain (decide-034, rev CLC-1.4)

Combined (11 vectors) # Scenario Expected Derivation C1 wildcard grant + params within bounds allow E2 + P1 C2 wildcard grant + params exceed bounds deny E2 + P2 C3 intersection + constraint violation deny I4 + D6 C4 two-source intersection, both narrow allow, narrowest I1 + I4 C5 grant with empty constraint bound deny deny-when-declared C6 unknown scheme in grant deny("unknown_constraint") fail-closed C7 grant: std/database-v1:query:*, op: std/database-v1:query:SELECT, params mismatch deny E2 + P4 C8 three-source intersection, one absent deny I3 C9 valid grant + valid constraint + valid op allow D1 C10 malformed id in operation deny("invalid_capability_id") D4 C11 delegation chain, intermediate hop declares empty bound deny deny-when-declared propagates Total: 154 vectors

• Decisions D17–D28 are the corpus pin for the residual-obligation channel unresolved / allow_unresolved, the Section 8.1 (scheme,type) identity, the invalid_constraint value grammar and the Section 9.1 multi-grant aggregation. decide-025/-026 and syntax-007/-008 pin the Section 3 scheme grammar; decide-021/-022, decide-019/-024 pin Section 8.1 time/network value shapes; decide-019 pins the no-cross-midnight rule; decide-028/-029/-030 pin Section 9.1 ({}≡absent, any-allow union, deterministic deny reason); decide-035 (D29) pins multi-grant residual-obligation union across covering grants; decide-031/-032/-033/-034 (D31–D34) pin the Section 8.1 max_rows request-side value domain (rev CLC-1.4); decide-018 (D30) pins Section 3 wildcard-shape detection ahead of the base grammar.

• The corpus additionally carries 6 scheme stress-test vectors (clinical-001/-002, payments-001/-002, data-001/-002) exercised against std/{clinical,payments,data}-v1: their enum/bound outcomes follow the grouping rules above (→ B.3 params semantics), and payments-002 additionally exercises Section 8 fail-closed unknown_constraint for a scheme-scoped constraint type. These vectors add no new normative rule; they exist to record the v2 requirements evidence, not to extend v1. undeclared-001/-002 (→ B.5 D13/D14) pin the Section 6.2 key-closure rule. params-006/params-013 (→ B.3 P6/P13) close two previously-unmapped rows; params-020/021 (→ P20/P21) are the positive boundary cases of params-018/019; decide-016/017 (→ D15/D16) pin the Section 9.1 pre-check and layer-1 paths for absent/empty grant and id-less operation; intersect-007..010 (→ I7..I10) pin Section 7 rules 5–6 including empty-params sources, order independence, and a params-free identifier comparison; intersect-011/-012/-013 (rev CLC-1.15) are the cross-type audit pins of Section 7 rule 6 (string × number merge refusal, type-sensitive enum-member equality, 1.0 ≡ 1) and sit outside the B.1–B.6 tables. The CLC-1.16 input-boundary pins (params-053–params-062) and the CLC-1.17 numeric-corner pins (params-063–params-070) are tabulated in B.3 (P53–P70 above), so the corpus total is 154. B.5 row labels Dn are semantic row numbers, not corpus ids (D15/D16 ↔ decide-016/-017, D29 ↔ decide-035); D10 (absent/empty grant, operation present) has no dedicated vector — the pre-check path is pinned by decide-016 (D15).

Containment (external corpus) The delegation-containment relation (Section 13) is pinned by a separate corpus at capability/data/_vectors/clc-d/: containment-vectors.json (64 vectors, groups: identifier coverage, parameter narrowing, constraint non-participation, validity, symmetry), containment-property-cases.json (784 forward-closure cases over 39 shared operations) and containment-crosswalk-vectors.json (45 cross-vendor vectors). It is not re-tabulated here; the corpus README maintains the live count and the snapshot date.

Extended parameter bounds (external corpus) The Section 6.5 param_bounds grammar is pinned by capability/data/_vectors/clc-v1/param-bounds-vectors.json (49 vectors, groups: numeric interval/step, enum cardinality, optional keys, nested recursion, the binding rule, malformed bounds, scheme defaults), with param-bounds-vectors.schema.json. It is not re-tabulated here. These vectors exercise a CLC-1.10 field; a CLC-1.9 or earlier implementation refuses them through the Section 12.1 minor gate rather than ignoring the bounds. The JSON type-sensitive enum equality of Section 6.5 layer 8 (rev CLC-1.15) is pinned by the companion file param-bounds-equality-vectors.json in the same directory.

Resolve (external corpus) The Section 8.5 Resolve function is pinned by capability/data/_vectors/clc-v1/resolve-vectors.json (26 vectors, groups: terminal pass-through, all-satisfied → allow, partial → allow_unresolved, violated → deny, the time:window core clock in and out of window, TTL re-evaluation, bare assertion without now, conflict precedence and malformed input), with resolve-vectors.schema.json. Vectors that exercise the core clock carry now; a CLC-1.10 implementation that does not implement Resolve is unaffected by this corpus.

ConstraintUnion (external corpus) The Section 7.1 derived ConstraintUnion projection is pinned by capability/data/_vectors/clc-v1/constraint-union-vectors.json (12 vectors: union over one/many grants, duplicate folding across sources, deterministic ordering, empty constraints, and the empty-chain absent_source refusal), with constraint-union-vectors.schema.json. The UTF-8 byte-order collation pinned in rev CLC-1.15 (Section 7.1) — where UTF-16 code-unit order would diverge — is exercised by the companion file constraint-union-collation-vectors.json in the same directory.

AuthorizeWithChain (external corpus) The Section 13.11 fused chain check is pinned by capability/data/_vectors/clc-d/authorize-chain-vectors.json (15 vectors: empty chain, broken identifier/parameter hop, a broken hop reported before op validation, the degenerate two-grant form, ancestor-constraint enforcement via the effective intersection, the param_bounds refusal, and pass-through of allow / allow_unresolved / operation-layer deny), with authorize-chain-vectors.schema.json.

BoundMeet (external corpus) The Section 6.6 intersection of param_bounds is pinned by capability/data/_vectors/clc-v1/param-bounds-meet-vectors.json (groups: numeric min/max/step meet including the coarser-grid and fail-closed step cases, enum intersection and tightened cardinality, optional conjunction, nested recursion, the empty meets (min>max, disjoint enums — empty within one value family), the unrepresentable crosses (numeric∩enum in either order and scalar∩nested, refused since rev CLC-1.15; cardinality-only enum, incommensurable steps → invalid_params_binding), the identity empty Bound, and the cross-site refusal), with param-bounds-meet-vectors.schema.json. The rev CLC-1.15 meet invariant — every successful meet authorizes only what every source authorizes — is exercised by param-bounds-meet-property-cases.json in the same directory.

Containment Carrier Vocabulary and Reason-Code Mapping This appendix is informative. It puts a carrier's own vocabulary and its failure codes beside CLC-D's, so an adopter can explain a containment verdict in the carrier's terms — and can see exactly where the mapping is many-to-one and therefore not reversible without carrier context. The profiles named here are the pinned ones in Section 13.9.3; the profile obligations are Section 13.8.1.

Vocabulary Carrier Carrier term CLC term used here Note CLC-D identifier scheme:action:path CapabilityId the relation's left/right identity CLC-D parameter bound params declared bound; absent or {} = unconstrained CLC-D constraint (varwof/constraint-v1:*) constraints union axis, outside Contains (Section 13.4.4) ATN resource_bounds, numeric conditions params (upper bounds) Section 9.2 takes the per-dimension min ATN id + schema.digest identifier segments digest mismatch ⇒ distinct capability (ccx-043) ATN preconditions — (union axis) never compared by Contains AAT tools(derived) ⊆ tools(parent) identifier coverage   AAT argument constraints (exact/range/one_of/…) params AAT's own contains/subset are constraint types, not the relation AIP scope, budget, expiry, domains identifier + params absent ⇒ nearest ancestor, resolved by the profile AAE mandate.actions identifier coverage   AAE CONSTRAINTS value (unwrapped) params numeric ≤, allowlist ⊆ AAE validity — (carrier/time, CLC-A R1) not a declared Contains input AOA operation scope string identifier path * maps to a CLC trailing wildcard AEGIS dotted capability identifier path   AEGIS numeric context params   AEGIS allowed_roles, environment, risk_level, scope — (not mapped) identity/policy dimensions with no CLC field; the profile leaves them to the carrier (C.3)

Failure-code mapping Carrier failure codes collapse onto CLC-D's three codes. The mapping is many-to-one and is therefore not reversible without the carrier recording which of its own codes applied before mapping. Carrier Carrier code / failure CLC-D reason ATN Section 9.1 identity rule different id different_namespace ATN Section 9.1 identity rule different schema.digest child_exceeds_parent ATN Section 9.2 raised resource_bounds params_not_narrower AAT I4 tool outside parent set different_namespace AAT I4 constraint not subsumed params_not_narrower AIP Section 4.4 aip_scope_insufficient child_exceeds_parent AIP Section 4.4 aip_budget_exceeded params_not_narrower AIP Section 4.4 aip_depth_exceeded child_exceeds_parent AAE Section 3 action not a subset different_namespace AAE Section 3/Section 7.4 constraint not more restrictive params_not_narrower AOA Section 6.2 scope not strictly narrower child_exceeds_parent AEGIS AIAM1-DEL-010 authority exceeds the delegator's params_not_narrower AEGIS AIAM1-DEL-011 capability the delegator lacks different_namespace The collapse is visible in the right column: child_exceeds_parent is reached by both AIP scope and AIP depth failures and by an ATN digest mismatch, and params_not_narrower by every carrier's bound-widening. A carrier that needs its own code back must keep it alongside the CLC-D reason; CLC-D guarantees the stable language-level reason, not the carrier's.

Profile obligations exercised by the corpus Two of the Section 13.8.1 obligations are visible in the pinned vectors: Identity dimensions the language cannot carry fail closed. ATN's schema digest is carried as an identifier segment (ccx-043) rather than dropped, so a digest mismatch is child_exceeds_parent instead of a silent match. Semantics the relation cannot see are a profile limitation, not a CLC gap. AEGIS's allowed_roles/environment/risk_level and AAE's validity have no CLC field; the profile leaves them to the carrier and asserts nothing about them. They are not modelled here, and a profile MUST NOT claim containment of a dimension the relation cannot see.

Acknowledgements Iman Schrock (EMILIA Protocol) reviewed the intersection and constraint semantics against the revision 1.1 corpus and supplied the adversarial cases that revisions 1.2 and 1.3 fix: nested partial overlap in intersection, constraint value handling for max_rows, and the public entry-point contract. For revision 1.4 he re-ran the Go, Python and TypeScript implementations and the 1,184 property cases, and closed the two objections he had raised against the max_rows value domain and the UTF-8 size bound. For revision 1.5 he re-ran the three Go suites — 105 authorization, 30 evidence and 13 crosswalk cases — and supplied the four corrections that revision carries: the collapse from three-valued evaluation to the binary report (Section 10), the separation of allow_unresolved from evidence (Section 11), the scope of delegation (Section 12), and the boundary between this projection identity and CAID (Sections 4.2, 4.3 and 6.4). For revisions 1.6 to 1.8 he re-ran the three implementations and reported the raw-boundary cases those revisions fix: the HTML-escaping and key-order defects in the canonical serializer, the decoded paths that disagreed with the raw path on malformed Unicode, and the raw size checks that counted a number by its received spelling and a literal astral character by UTF-16 code unit. Revision 1.9 folds the delegation-containment relation and the CLC-D class into this document (Section 13, Appendix C), retiring the formerly separate containment extension, and pins it against six adjacent capability drafts (ATN, AAT, AIP, AAE, AOA, AEGIS) reviewed on 2026-09-21.