Governance, skills, and hooks stack for running Claude (or any capable LLM) as an autonomous senior engineer on any software project — frontend or backend, any language. MIT.
LoomA8osAgent/a8-loom-coordinator
Agent
WHAT THIS IS: the AUDITOR agent archetype — sweeps for duplication / orphans / dead code. Its findings are UNTRUSTED until grep-confirmed live. It embeds the audit-agent contract so a "duplication" finding can't cause a wrong move (dead code gets deleted, not extracted).
LoomA8osAgent/a8-loom-coordinator
Agent
WHAT THIS IS: the BUILDER agent archetype — executes build/edit work on its assigned files. It boots with the domain canon preloaded so it EXTENDS shared homes instead of re-rolling per-module copies. It NEVER touches git — it edits files, returns its diff + a summary, and the coordinator commits.
LoomA8osAgent/a8-loom-coordinator
Agent
WHAT THIS IS: the PLANNER agent archetype — the governance-aware alternative to a built-in Plan/Explore agent (those skip CLAUDE.md). It investigates + designs + authors specs, but writes NO application code. A builder executes its plan.