| name | ch-3-test-auto |
| description | Runs the automated test phase loop for the current feature branch by invoking ch_3_test_auto.py. Writes failing tests for all [TEST] tasks, runs the iterative test-critic and test-quality-review two-gate loop, applies fixes, and escalates on failure. Run manually after reviewing tasks.md. |
| user-invocable | true |
Test Auto-Orchestrator
Run the automated test phase loop for the current feature branch.
All orchestration logic lives in .claude/agents/ch_3_test_auto.py. This skill is a thin
invocation wrapper — do not re-implement the loop here.
Execution
Run from the repo root:
python .claude/agents/ch_3_test_auto.py
The script derives the feature from the current git branch automatically.
To target a specific feature, pass --feature <name>:
python .claude/agents/ch_3_test_auto.py --feature 015-job-description-rich-text
Wait for the script to complete and relay its output to the user.
What the script does
- Validates pre-flight conditions (spec.md, plan.md, tasks.md exist)
- Runs the test agent to write failing tests for all unchecked [TEST] tasks in tasks.md
(skipped if all [TEST] tasks are already checked off)
- Runs an iterative two-gate review loop (up to 3 iterations):
- Gate 1 — Test critic: validates task traceability, red-state confirmation, spec
coverage, implementation leakage, test isolation, and stack compliance
- Gate 2 — Test quality review: validates assertion quality, test naming, and CI
readiness (runs only after Gate 1 passes)
- Either gate FAIL → fix agent addresses violations in test files only → both gates re-run
- Both gates PASS in the same iteration → done
- On PASS: triggers auto-commit via the
after_test git extension hook
- On 3-iteration exhaustion: writes
specs/$FEATURE/ch-3-test-critic-escalation.md and exits non-zero
The script is resume-safe: re-running after any interruption continues from the last
incomplete step using result files as idempotency markers.
What this skill does not do
- Does not implement any feature logic — that is done by
/speckit-implement or /ch-4-implement-auto
- Does not push to remote or open a pull request
- Does not run the implement phase — run
/ch-4-implement-auto after reviewing the test files
