Skip to main content

Workflow tools & project intelligence

Individual tools are primitives. These chain them into real, verified outcomes — this is the "AI employee" layer, in tools/workflowOps.js and tools/projectOps.js.

Workflow tools​

ship_change​

Inspects your current changes, creates or checks out a branch, runs your project's checks, commits, pushes, and opens the PR — one call instead of eight.

fix_ci + verify_ci_fix​

fix_ci finds the latest failing GitHub Actions run for a branch, pulls the real job logs (not just "it failed"), and returns them for diagnosis. Once the underlying code is fixed, verify_ci_fix commits, pushes, and polls until the run is actually green — not just "pushed and hoped."

deploy_project​

Checks project health, runs tests and the build, deploys, polls for completion, and then HTTP-verifies the live URL is actually responding — not just "deployment triggered."

Project intelligence​

Auto-detects your stack instead of guessing generic commands:

  • project_detect — language, framework, package manager
  • project_info — structured summary of the above, plus resolved test/lint/build commands
  • project_health — git cleanliness, ahead/behind, dependency status
  • run_tests, run_lint, run_typecheck, run_build — run the correct command for the detected stack (npm test vs pytest vs cargo test, etc.) rather than assuming one ecosystem

Example​

"Ship this change."

Claude calls project_health → run_tests/run_lint → ship_change (confirm: true) — branch, commit, push, PR, in one guided sequence instead of Claude re-deriving the steps each time.