karvey-iteratelisted

# Karvey Iterate — The Iteration Engine > **Afán** = don't stop until the result is really achieved. This skill is where that lives: it takes > what testing/QA/real-runtime surfaced and sends each item back to where it belongs, instead of > letting it die at the end of a linear pipeline. ## Purpose The pipeline (0→12) is the happy path. `karvey-iterate` is the **feedback brain** that closes the loop. It does ONE thing: read the findings inbox and **route** each finding to its correct edge. The phase skills (`test`, `qa`, `browse`) only **observe and classify**; this skill **routes**. Keeping routing in one place is the whole design. Read these rules before acting: - `karvey/rules/iteration-loop.md` — the three feedback edges, `findings.md`, the spec-revision sub-cycle, convergence. - `karvey/rules/incident-tracking.md` — the `BUG-NN` tracker with state history. - `karvey/rules/backlog.md` — the dual (Markdown + ClickUp) discovery backlog. - `karvey/rules/phase-close.md` — the close ritual. ## When to run it - After `test` / `qa` / `browse` produced findings. - Any time a finding appears mid-flow ("this spec was wrong", "a new bug showed up", "good idea but out of scope"). - The orchestrator routes you here whenever `findings.md` has `open` items. ## Execution steps ### Step 1 — Load context Read: - `docs/spec/changes/{change-id}/spec.json` - `docs/spec/changes/{change-id}/findings.md` (the inbox; if it doesn't exist, there's nothing to iterate — tell the user and sto