Intent Route Resolver

SkillDev tools

Resolve a normalized intent against one digest-bound finite route catalog and return a non-authorizing candidate, ambiguity, no-match, or invalid disposition.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Intent Route Resolver skill

What this skill tells your AI

The instructions your AI receives, as published by cyberalchemyai/arcanum in arcana/intent-route-resolver/SKILL.md and read by ahel’s review.

intent-route.request@1 + digest-bound catalog -> candidate | ambiguous | no-match | invalid

The result is advice about possible routing only. It never grants permission, opens work, binds a workflow, schedules execution, or records effects.

  • one normalized intent-route.request@1 value;
  • one finite intent-route.catalog@1 value whose content digest equals the request binding;
  • an adapter-provided capability token for the local JSON port.
  1. Reject unsupported protocol, core, manifest, or content-digest bindings.
  2. Validate request and catalog as closed values.
  3. Preserve every discriminator posture: declared, inferred, or unresolved; never fill an unresolved value.
  4. Evaluate exclusions before requirements.
  5. Use only catalog-declared dominance to remove eligible alternatives.
  6. Return ambiguous when missing information can change eligibility, candidate only for one non-dominated eligible route, no-match without fallback, or invalid for admitted semantic input failures.
  7. Emit one canonical intent-route.runtime-port@1 response with a complete route trace and authority_effect: none.

The resolver has no discovery, repair, ranking, registration, approval, authorization, persistence, dispatch, scheduling, execution, host-spawn, network, filesystem-write, or effect authority. A consumer may display the result, ask the exact clarification, collect a replacement request, or stop. Any later binding or execution remains owned by the consumer's existing governed workflow.

  • all protocol and digest bindings are explicit and fail closed;
  • catalog order cannot change canonical response bytes;
  • removing evidence cannot improve a disposition;
  • every evaluated route appears in the trace;
  • transport rejection is never presented as a trusted disposition;
  • candidate source and generated closures contain no product-specific imports;
  • the manifest declares zero permissions, zero side effects, and no authority.
  • inferring missing discriminators;
  • semantic retry, compatibility fallback, or model escalation;
  • treating one candidate as approval;
  • creating a route registry or a second decision lifecycle;
  • importing a consumer runtime, product model, or private implementation;
  • claiming product acceptance from Node or browser portability evidence.

Return exactly one canonical JSON document matching either intent-route.runtime-port@1 or intent-route.error@1, with the documented numeric exit code and zero stderr bytes.

Signals

GitHub stars
25
Forks
3
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
intent-route-resolver
Source
github.com/cyberalchemyai/arcanum