multi-agent-governancelisted

# Multi-Agent Governance (Zero-Runtime) Govern multiple AI coding agents working on the same GitHub repository **without a daemon or platform dependency**. Everything runs on native GitHub mechanisms: issues as the task market, assignees as claims, PR + CI as the merge gate, auto-merge as the fully-automatic pipeline. ## When to use - Multiple AI sessions will work on the same repository in parallel. - You want CI-enforced PR gates (identity, diff hygiene, docs sync, conflict guard) without installing anything that must stay running. - A single agent needs to know the repo governance protocol before opening a PR. ## Core protocol (six gates) 1. **Identity** — every commit carries a session key trailer; missing identity fails the gate. 2. **Claim** — work is claimed on an issue before a branch is opened (issue = task market). 3. **Diff hygiene** — generated files, secrets, and protected paths are blocked by path rules. 4. **Docs sync** — documentation changes must pass sync/consistency checks. 5. **Conflict guard** — overlapping critical paths with another open PR block the merge. 6. **Merge** — green CI + fast-lane rules decide auto-merge. ## Steps 1. Read the repo root `AGENTS.md` (the single thin entry point generated by the scaffold). 2. Pick an open issue, claim it (assign yourself), and comment with a plan. 3. Create a branch `codex/<slug>`, implement, and open a PR that references the issue. 4. Let CI run the gates; fix any failure locally and re-push (max 3 retr