← ClaudeAtlas

verifylisted

Use after editing code in a moon monorepo, before commit or PR, to verify the change — runs affected :typecheck/:lint/:test via the verify-runner subagent, dispatches change-auditor for an adversarial review of the diff against scope, reports evidence (not assertion), and loops back on a torn wing (failing check). Prefer this over a blind repo-wide test run.
g-bastianelli/nuthouse · ★ 0 · Code & Development · score 65
Install: claude install-skill g-bastianelli/nuthouse
# verify > At visible transitions, try to dispatch `warden:voice` with `SUMMARY: <≤15 words, in the user's language>`, `PERSONA_CONTRACT_PATH: ${CLAUDE_PLUGIN_ROOT}/shared/persona-line-contract.md`, and `VOICE_FLAG_PATH: $HOME/.claude/nuthouse/voice.state`. Print the returned `line` only when non-empty. If `warden` is unavailable, errors, returns malformed output, or voice is disabled, print nothing and continue. Never make voice a precondition, never retry, never mention missing `warden`. > Autopilot flag: !`cat "$(git rev-parse --path-format=absolute --git-common-dir 2>/dev/null)/nuthouse/autopilot.json" 2>/dev/null || echo off` — monkey-maestro relay. Take the **autopilot branch** in Steps 3–4 ONLY if this resolved to JSON with `active: true` and a future `expires_at` (the flag lives under this repo's shared `.git`, so it is already repo-scoped); otherwise behave interactively as usual. ## Voice Read `../../persona.md` at the start of this skill. The moon-moth voice is canonical for wrapper lines; the report stays plain. A clean flight (🌙) is declared only on real passing output. **Scope:** local to this skill's execution only; revert to the session default voice after the final report. This skill is **rigid** — execute steps in order. ## Language Adapt chat output to the user's language. Project ids, task names, file paths, and CLI flags stay in their original form. ## When you're invoked A change is made and needs verifying before it ships. The moon-moth checks