Multi-runtime AI agent governance harness — Claude Code · Codex · Gemini under one YAML policy, with portable security hooks
Skill Claude CodeCodex
Distills quarantined raw/ session captures into curated typed atomic notes in the brain's notes/, enforcing dedup, >=2 typed edges, provenance, and a lint=0 gate, then promotes evergreen notes to the human vault via /wiki-ingest. NOT for hand-writing straight into notes/ or the vault (that bypasses provenance + the…
Skill Claude CodeCodex
Multi-vendor code review council — runs the Claude code-reviewer agent alongside external second opinions (codex, gemini) in parallel, then synthesizes with citation verification, source tagging, and disagreement surfacing. Optionally seats grok (--with-grok) and/or an OpenRouter free-tier advisor (--with-free) as…
Skill Claude CodeCodex
Agent-driven, read-only self-audit of the harness — run the machine integrity layer once as a dry-run, present a per-check pass/fail table, cite the P1-1 doc-reality result, and for any failure give root-cause + fix + backlog follow-up. Consumes the machine gates; it does not reimplement them. NOT for auditing a…
Skill Claude CodeCodex
Router over the harness skills — which skill fits your situation, and the main flow through them. NOT for performing the routed work itself — it only points; invoke the skill it names.
Skill Claude CodeCodex
The §5 autonomous-improvement-loop procedure specialized to this harness's own reviewer prompts (agents/code-reviewer.md + agents/security-reviewer.md) — the mission-specific 9-step regulation layered on skills/loop's generic mechanics. NOT for an arbitrary mission (use /loop for that), and NOT for editing…
Skill Claude CodeCodex
Run a mission as a bounded series of fresh-context, one-task-per-turn attempts against core/infra/loop-run.sh — hard attempt cap, per-attempt timeout, and a circuit-breaker on repeated GATE failures, resumable after a restart from on-disk state. NOT for a single one-off task (just do it directly), and NOT for…
Skill Claude CodeCodex
Meta-audit of a /supervise run — did the supervisor do its job? Runs the four-lane machine layer (core/infra/manager-audit.sh) over the run's logs, interprets the semantic halves, and turns actionable findings into concrete patch proposals in PROPOSALS.md for one-click user approval. Read-only analysis; NEVER applies…
Skill Claude CodeCodex
Seat a panel of real-distribution-grounded Korean citizen personas in front of a piece of UX, copy, or content and report how ordinary users would react. Samples 5 personas from a catalog, dispatches each as an independent panelist, and synthesizes their reactions. NOT a code/security/performance review (those are…
Skill Claude CodeCodex
Upstream planning-discipline — brainstorm a feature into spec.md + plan.md under .agent/plans/ /, then route to plan approval. Enforced by the spec-gate tool boundary, not prompt coercion. NOT for trivial/test/doc-only edits (the gate scopes those out), and NOT for executing an already-approved plan (that is…
Skill Claude CodeCodex
Dispatch a multi-wave plan to specialist agents with audit + risk-area abort. Supports --auto-push, --auto-merge, --goal-mode for budgeted runs, --verify-blocking for a hard completion-claim gate. NOT for writing the plan itself (that is /spec), and NOT for a single small edit with no waves — just make the edit.
Skill Claude CodeCodex
Independently re-verify a completion claim in a separate context — mechanical evidence (files/tests/assertions) plus a refute-by-default semantic judge — before a wave or task is allowed to be called "done". Emits the shared verdict schema. NOT for code-style review (that is a reviewer's lane) and NOT for verifying…
Skill Claude CodeCodex
Per-lane install → auth → verify onboarding for the cross-vendor worker lanes (codex, antigravity/gemini, grok, kiro) with a cost-model/tier briefing before anything is installed. NOT a dispatcher (core/infra/call-worker.sh is), and NOT a paid probe run without explicit user approval — every real round-trip probe is…
Skill Claude CodeCodex
Commits staged changes and opens a PR. Runs gitleaks + risk-area guards before committing. Push is user-confirmed by default; --auto-push and --auto-merge are opt-in. NOT for review or verification (run those before wrapping), and NOT when nothing has changed yet — there is nothing to commit.