Skip to content

Business operating system for Claude Code — 57 skills, 21 agents, smart daemon. Unified inbox (WhatsApp/Email/Slack/Telegram), autonomous PR merge, full-AWS monitoring, revenue (Stripe+RevenueCat), e-commerce (Shopify), marketing (Klaviyo/Meta/GA4), voice (Bland/ElevenLabs), APM (Datadog/NewRelic/OTEL), YOLO mode.

3.2 KB
Raw Blame

claude-ops — default approval policy

Repository-wide rules for Cursor Approval Agents evaluating pull requests in this repo.

Purpose

claude-ops is a business operating system plugin for Claude Code. Changes can affect safety hooks, deploy automation, credential handling, and fleet orchestration. Default posture: approve only when risk is low and automated review is clean.

Auto-approve when ALL are true

  • PR is small and scoped (generally ≤ 300 changed lines, excluding lockfiles and generated docs)
  • Bugbot Review Context reports no findings requiring human review
  • Security Review Context reports no findings requiring human review (when enabled)
  • Risk score is at or below the agent's configured maximum threshold
  • CI checks required for the changed paths are green
  • No changes under high-risk paths (see below) unless a more specific subtree policy explicitly allows it
  • PR does not modify approval policy files, routing files, or safety-hook bypass logic

Safe auto-approve examples

  • Documentation-only edits under claude-ops/docs/**, docs/**, or README.md
  • Test-only changes that do not alter production hook or deploy behavior
  • Changelog, version bump, or badge updates with no runtime logic changes
  • Typo or copy fixes in skills where behavior is unchanged

Never auto-approve

  • Any change to hooks/**, claude-ops/hooks/**, or safety-hook configuration
  • Credential, secret, rotation, or auth flows (**/credentials/**, **/rotate/**, Doppler/1Password wiring)
  • MCP server implementations (claude-ops/mcp-servers/**) that add network calls, file access, or new tool surfaces
  • Launchd/daemon lifecycle changes (claude-ops/launchd/**, claude-ops/scripts/** that start/stop daemons)
  • GitHub Actions workflow changes (.github/workflows/**) that weaken CI, skip tests, or broaden deploy permissions
  • Deletions or relaxations of guardrails (rm-rf blocks, secret-scan hooks, main-branch push warnings)
  • Changes that remove tests, disable lint/typecheck, or bypass pre-commit hooks
  • PRs labeled security, breaking, or do-not-auto-approve

Reviewer routing

When auto-approval is not allowed, request reviewers based on changed area:

Changed area Reviewers
Hooks / safety / deploy auto-fix @Lifecycle-Innovations-Limited/platform or repo maintainers
MCP servers / external integrations @Lifecycle-Innovations-Limited/platform
Plugin manifest / marketplace metadata @Lifecycle-Innovations-Limited/platform
Docs-only Any maintainer; no specialist required

If reviewer assignment is unavailable, leave the PR unapproved and add a comment summarizing why human review is required.

Deploy expectations

For changes that ship through dev → staging → main:

  • Do not auto-approve solely because CI passed on the PR branch
  • Require evidence that deploy-fix and health-check paths remain intact when touching deploy-related code
  • Treat production-impacting script changes as medium risk minimum

Conflict resolution

If this file conflicts with a closer APPROVAL_POLICY.md or a routed policy in .cursor/approval-policies/, follow the more specific policy. If specificity is unclear, follow the stricter rule and do not auto-approve.