skills.sh
AvailableUniversal skill installation, including Claude Code and Codex.
npx skills add https://github.com/ulpi-io/skills-autonomous-engineering --skill auto-testauto-test
Builds a green, meaningful suite whose new tests are proven capable of detecting broken behavior.
WORKFLOW / 05/auto-test [scope — path/module/diff, or 'flaky' to stabilize] (default: the current diff)
- 01
Identify the real test command, baseline signal, scope, and run budget.
- 02
Discover untested behaviors rather than chasing a percentage.
- 03
Write one behavior test, break the implementation, and require the test to fail.
- 04
Restore the implementation and converge the suite to green.
- 05
Diagnose flakes at their root and report all remaining failures honestly.
▸ operating envelope
Use it deliberately.
The shortest useful definition of where this skill helps—and where it should stay out of the way.
Use when
- Implemented behavior needs coverage and a trustworthy green signal.
- A risky module or known bug needs regression tests.
- A flaky suite needs root-cause stabilization rather than retries or bigger timeouts.
Do not use when
- You are implementing new behavior test-first; that belongs inside auto-build.
- The repository has no test runner configured.
- Expected behavior is a product decision that has not been resolved.
▸ output contract
What must be true when it stops.
Produces
- Behavior-focused, mutation-verified tests
- Baseline-to-final suite and coverage signal
- Real-defect and flake-root-cause ledger
- .ulpi/runs/<id>.json checkpoint
Guarantees
- Every added test is mutation-checked before it counts.
- The skill does not delete, skip, isolate, suppress, or weaken tests to manufacture green.
- A test exposing a real product defect is surfaced rather than rewritten around the defect.
- A capped red run reports converged:false with the exact failing units.
- The checkpoint makes broad coverage work resumable.
Honest failure states
- converged:false — the suite remains red when the cap or budget fires
- tautology_rejected — the test stayed green against a broken implementation
- bug_surfaced — the test is sound and the implementation is wrong
- blocked_ambiguity — correct expected behavior needs a human decision
▸ enforcement
Know which claims are executable.
Scripts are linked when this skill owns deterministic machinery. The remaining rules are explicit operating contracts and must not be marketed as hooks.
▸ install and invoke
One skill, three honest surfaces.
Claude Code plugin
AvailableInstall the full plugin, then invoke /auto-test or let routing select it from context.
Codex plugin
Work in progressA work in progress on the codex-native-plugin branch — build it, start a new session, then invoke $autonomous-engineering:auto-test.
▸ composes with