autorun-dryrun

name	autorun-dryrun
description	Run the autorun pipeline in DRY_RUN mode against a fixture spec and assert each expected artifact landed. Use as a smoke test after editing scripts/autorun/* to confirm orchestration wiring still works without burning API cost.
disable-model-invocation	true

/autorun-dryrun

Smoke-test the MonsterFlow autorun pipeline in DRY_RUN mode.

What it does

1. Stages a fixture spec (tests/fixtures/autorun-dryrun/sample.spec.md) into a temporary queue/ directory under $TMPDIR.
2. Runs AUTORUN_DRY_RUN=1 bash scripts/autorun/run.sh against that queue.
3. Asserts every expected artifact landed in queue/sample/:
   • review-findings.md (or stub equivalent)
   • risk-findings.md
   • plan.md
   • check.md
   • pre-build-sha.txt
   • build-log.md
   • verify-gaps.md containing VERDICT: COMPLIANT
   • state.json
4. Cleans up the temp queue.
5. Reports pass/fail summary with diffs to expected artifact list.

Execution

Invoke via the user's terminal (not in-session — the skill runs the actual pipeline):

bash tests/autorun-dryrun.sh

The runner script (tests/autorun-dryrun.sh) is the canonical implementation. This SKILL.md exists so the user can invoke /autorun-dryrun and Claude knows to execute that script and surface the output.

When to invoke

• After any edit to scripts/autorun/*.sh — confirms orchestration still works end-to-end
• Before tagging a release — final smoke test
• When debugging stage handoff issues — isolates whether the wiring is broken vs the stage logic

Output

Pass:

✓ all 8 artifacts present
✓ verify-gaps.md contains VERDICT: COMPLIANT
✓ state.json valid
PASS — autorun dry-run complete in 3.2s

Fail:

✗ missing: queue/sample/check.md
✗ verify-gaps.md does not contain VERDICT: COMPLIANT (got: VERDICT: INCOMPLETE)
FAIL — 2 assertions failed

Implementation note

This skill is disable-model-invocation: true because it has side effects (writes/deletes a temp queue, executes shell pipeline). Only the user can fire it via /autorun-dryrun.