- Problem: the registry fixtures validated SKILL.md against the committed docs/COMMAND-REGISTRY.json but nothing in the file asserted the registry itself was fresh, so a standalone run passed against a stale registry (npm test does run command-registry:check first, but this file is also run on its own).
- Fix: rebuild the registry in memory via the exported generateRegistry() and assert deep equality with the committed JSON so drift fails as a regenerate hint with a field diff instead of passing silently; add one sentence to SKILL.md stating that a command retired from commands/ fails closed with the BLOCKED marker like a no-argument command.
- Verification: node tests/skills/plan-orchestrate-registry.test.js passes with the new check 8 (8/8); sabotage with a mutated registry description fails check 8 with the regenerate hint and a field diff.
- Problem: the emitted command does not consume the advertised agent chain
(each command owns its own execution behavior), and steps mapped to
argument-less commands were emitted as bare commands with no
machine-readable link to the intended step.
- Fix: adopt the small-router role — emit only commands whose documented
argument form can carry the [Plan: <path>#step-<id>] scope marker
(orch-*, /plan), fail closed with a BLOCKED entry and a manual-use note
for every other command, derive quoted argument forms from the command
files instead of an authoritative hand-written table, and drop the
agent-chain machinery entirely. Adds
tests/skills/plan-orchestrate-registry.test.js, which validates every
command the skill names against commands/<name>.md and
docs/COMMAND-REGISTRY.json and fails when they drift.
- Verification: node tests/skills/plan-orchestrate-registry.test.js (7
fixture groups pass; sabotage-checked against chain reintroduction,
wrong scope-carrying claim, and retired-command reference).
Santhi Prakash