local-agent-harness

@grammy-jiang

Manage the local AI-coding-agent harness for any repository — empty, partially configured, or fully mature. Detects the repo's maturity stage (S0 greenfield → S3 production), audits existing manifests for drift (missing sections, relaxed constraints, stale overlays), and applies a stage-appropriate harness covering both the agent (AGENTS.md with inlined HC1–HC6 hard constraints, plan.md, tool DAG, permission ladder, governed memory, cost/context budgets, runtime overlays for Claude Code / Codex CLI / Copilot CLI) and the repo (sandbox/devcontainer, pre-commit, verify CI on GitHub Actions, governance CI, secrets/SAST/dep scans, AI-readiness score). Use when the user says "make this repo agent-ready", "bootstrap AGENTS.md", "audit our agent setup", "score AI-readiness", "refresh the harness", "check whether AGENTS.md is up to date", or "evolve the harness as the repo grows". Three modes — check (read-only audit), init (render missing artifacts; never overwrites), refresh (back up + rewrite stale artifacts).