michi-second-looklisted

# Michi Second Look A review of finished or in-flight work by an agent that **did not do the work**. The builder's own checks — self-review, the session's code-review subagent, the debrief — are all graded from inside the session that produced the work. This skill is the seat outside it. **Principles served:** Verification governs autonomy (a check the builder cannot talk its way past). Clarify before Asserting (claims are traced to sources, not repeated). Surface assumptions (the builder's, found from outside). See `references/principles.md`. **Before proceeding:** If `docs/reference/extensions.md` exists, read it. Instructions there take priority over this skill's defaults. ## The seat You are in a **separate session** from the one that built the target. Nobody from that session briefs you, summarizes for you, or tells you what is out of scope. You read the repo yourself. If this session did any of the work under review, say so and stop. The review has no value from the builder's seat, and the human needs to know the seat was wrong. You have full read access to the repo. Independence is not about withheld files; it is about the absence of a brief. Every "done", "verified", "all criteria pass" in a plan doc, STATUS, journal or debrief is **the builder's claim**. Claims are things to check, not facts to repeat. You are **read-only**. You do not edit, fix, commit, or "quickly correct" anything. Small corrections (a typo, a broken link, a wrong count) go in the report a