Skip to main content
Use this checklist before enabling new OpenClaw capabilities or expanding autonomy.

​
Gate A: Documentation

  1. Integration plan is current: internal/openclaw-integration-plan
  2. Manager/support model is current: internal/manager-support-model
  3. Change protocol includes rollback path: internal/change-protocol
  4. Skill stack matrix is current: internal/skill-stack-matrix
  5. Known problems catalog is current: operations/known-problems-catalog

​
Gate B: Safety Controls

  1. Agent writes are limited to approved paths by policy
  2. Destructive operations require explicit human approval
  3. Every agent mutation has audit logs with actor and timestamp
  4. Rollback instructions are documented and tested
  5. Secrets are stored in Keychain/secret manager, not plaintext files

​
Gate C: Platform Health

  1. pnpm check passes
  2. pnpm format:check passes
  3. pnpm test:unit passes
  4. pnpm test:integration passes
  5. pnpm docs:check passes
  6. pnpm openclaw:check passes
  7. Render /api/health is healthy on active domains

​
Gate D: Routing and Responsibility

  1. Manager and support routing matrix is current
  2. Escalation paths are explicit for auth/data integrity/security incidents
  3. SLA and ownership for incident response are defined

​
Gate E: Channel and Node Policy

  1. WhatsApp external + Telegram internal split is documented
  2. External communications use neutral role language
  3. Dedicated VPS node is primary and MacBook node is fallback
  4. Runtime execution is denied by default
  5. One-off exec requires explicit manager confirmation

​
Gate F: SEO and Public Surface

  1. Canonical host points to active app domain
  2. robots.txt, sitemap.xml, and llms.txt align with real public routes
  3. Private/authenticated routes are not indexable

​
Gate G: Learning Loop

  1. Learning-loop policy doc is current
  2. Interaction log file exists and uses structured events
  3. pnpm openclaw:learn successfully updates operating profile
  4. Any promotion to autonomous behavior has manager signoff

​
CI Enforcement

This repository enforces a machine-readable readiness baseline:
  • script: scripts/check-openclaw-readiness.mjs
  • command: pnpm openclaw:check
  • CI step: OpenClaw readiness check in .github/workflows/ci.yml

​
Rollout Policy

  1. Start with supervised mode (advisory + PR generation)
  2. Enable narrow write scopes after one stable week
  3. Expand capabilities only with zero unresolved P1/P2 safety findings