AI agent commands
Canonical vocabulary for any AI agent (Cursor, Codex, Claude, Copilot, Windsurf, …).
Init (read first): AGENT_INIT.md
Rule file: .cursor/rules/ai-agent-commands.mdc
---
Commands
| You say | Agent does |
|---|---|
| Update? | Snapshot only — refresh master tasks + CI; no coding |
| next | Implement top P0/P1 open task from registry + GitHub issue |
| sync issues / sync tasks | npm run sync:issues (labels + tasks + project board) |
| autopilot | Triage PR checks + review threads until merge-ready |
---
Onboarding files
AGENT_INIT.md— all AI tools, read firstMISSION-CONTROL-WALKTHROUGH.mdTASK-TRACKING.mdplan/mission-control-master-tasks.md
---
Data hygiene
- Never commit or paste secrets
- Use cred vault for tokens
- Confirm before destructive actions (
confirmActionshared helper) - Procedure files track progress — not full chat transcripts
See Ecosystem Roadmap for floating AI and notification plans.