review-pipeline-logslisted

# Review Pipeline Logs **Announce at start:** "I'm using the bridge:review-pipeline-logs skill to check the pipeline's logs against the designed flow." ## Purpose Code that compiles and a pipeline that ran are two different claims. This skill closes that gap: it takes the design doc that says what the pipeline is *supposed* to do, takes the log output from an actual run, and checks one against the other — step by step, not just "did it crash." A pipeline that finishes with no errors can still have skipped a step silently or written a wrong output value; a pipeline that logs nothing useful will bite the next person who has to debug it in production. Both are failures this skill should catch. ## Step 1 — Locate the Design/Plan Doc This is the source of truth for what "correct" means. Search in this priority order: 1. A file path given in the user's request (if they named one) 2. `docs/plan.md`, `docs/spec.md`, `plan.md`, `spec.md` 3. Most recently modified `.md` under `docs/superpowers/input/` (gstack-to-plan handoffs live here) 4. Most recently modified `.md` under `docs/` (`ls -t docs/**/*.md 2>/dev/null | head -1`) 5. Most recently modified `.md` under `.gstack/` Read the located file in full. If nothing is found, ask the user which doc describes the pipeline's intended steps — do not guess a design from the log alone, since that would just be checking the log against itself. ## Step 2 — Extract the Expected Steps From the design doc, build an ordered list of pipeli