autoreview
01Command
One-shot autonomous review of a PR. Fetches review comments from AI bots, applies verified fixes, validates, and merges when safe.
Command
One-shot autonomous review of a PR. Fetches review comments from AI bots, applies verified fixes, validates, and merges when safe.
Command
Deeply inspect the agent-loop state file. View cycle history, thread details, errors, and state integrity. Useful for debugging and understanding loop status.
Command
Clear the agent-loop state and start fresh. Creates a backup of the current state before clearing. Use this when you want to reset metrics and start a new loop session.
Command
Restore agent-loop state from a backup. List available backups, restore from a specific backup, or use the most recent backup to recover from corruption or accidents.
Command
Resume the agent-loop from the last persisted state in .agent-loop/state.json. Loads cycle context and thread history, then continues the maintenance cycle.
Command
Start the agent-loop — continuous overnight/day repo maintenance cycle. Auto-detects existing state and prompts to resume or start fresh. Wakes every 5 minutes, triages repos, dispatches work to parallel agent threads, monitors progress, and reports results.
Command
Check agent-loop status — running state, cycle count, items processed, active threads, and recent results.
Command
Stop the agent-loop gracefully. Lets active threads finish within timeout, persists final state, and reports cycle summary.
Command
Run one-shot triage on repo queues. Scans open issues, PRs, CI status. Classifies items by fit, risk, autonomy. No loop, just snapshot.
Command
Migrate the current project's LLM integration to AnyRouter. Detect existing OpenAI / Anthropic / OpenRouter usage, swap base URL + key + model ids, preserve behavior, and add a smoke test.
Command
Create or update the project-local macOS buildandrun.sh script, wire the Codex app Run button, then use that script as the default build/run entrypoint.
Command
Inspect a macOS signing or entitlement failure and explain the minimum fix path.
Command
Run the smallest meaningful macOS test scope first and explain failures by category.
Command
Dispatch one or more headless Command Code (cmd) workers to implement well-scoped tasks or GitHub issues in parallel — each in its own git worktree, running cmd -p --yolo --max-turns 100000, committed to its own branch for review.
Command
Create a git commit with semantic commit message format and create a pull request.