qode-check

name	qode-check
description	Run qode quality gates such as tests and lint checks.

Quality Gates — qode

Run quality gates interactively in two sequential phases.

Phase 1 — Unit tests

1. Inspect the project structure to determine the test runner:
   • go.mod present → run: go test ./...
   • package.json with a test script → run: npm test (or yarn test if yarn.lock is present)
   • pytest.ini / pyproject.toml with pytest config → run: pytest
   • For other stacks, apply equivalent discovery
   • If multiple layers are present, run tests for each layer
2. Run all unit tests. Do NOT read qode.yaml — determine the command from the project structure only.
3. If any tests fail:
   • Stop. Do not proceed to Phase 2.
   • Output a structured summary of all failures across all layers:
     • Which tests failed and why
     • Proposed fix for each failure
   • Ask the user how to proceed with exactly three options:
     • Accept — apply the proposed fixes and re-run the failing tests
     • Stop — exit without making any changes
     • Comment — let the user add notes or corrections before retrying
   • On Accept: apply fixes, re-run Phase 1 (do not advance to Phase 2 until all tests pass)
   • On Comment: incorporate the user's feedback before retrying; do not loop blindly
4. If no test files are found: skip Phase 1, note the skip, and proceed to Phase 2.
5. If all tests pass: proceed to Phase 2.

Phase 2 — Linter

1. Inspect the project structure to determine the linter:
   • go.mod present and (.golangci.yml exists or golangci-lint is in PATH) → run: golangci-lint run
   • package.json with eslint in devDependencies or .eslintrc* file present → run: npx eslint .
   • ruff.toml or pyproject.toml with ruff config → run: ruff check .
   • For other stacks, apply equivalent discovery
   • If multiple layers are present, run the linter for each layer
2. Run the linter. Do NOT read qode.yaml — determine the command from the project structure only.
3. If linting issues are found:
   • Stop.
   • Output a structured summary of all violations across all layers:
     • Which rules were violated and where
     • Proposed fix for each violation
   • Ask the user how to proceed with exactly three options: Accept / Stop / Comment
   • On Accept: apply fixes, re-run Phase 2 (do not advance until lint is clean)
   • On Comment: incorporate the user's feedback before retrying; do not loop blindly
4. If no linter is found: skip Phase 2, report success.
5. If lint is clean: report that all quality gates passed and suggest running the qode-review-code step.

Unknown stack

If neither a test runner nor a linter can be determined for a given layer, report what was inspected and ask the user to specify the command. Do not guess.