reviewer-baselinelisted

# Reviewer Baseline Protocol This skill defines the shared review protocol that ALL code reviewer agents follow. It ensures consistent confidence scoring, finding classification, and output format across the review system. ## Output Mode Read this first — it governs HOW you record everything below. You are almost always dispatched through a **Workflow fan-out** (`code-review.mjs` for Stage 8 and standalone `/review-lead`; `intake-review.mjs` for intake) that hands you a **JSON schema**. When you have a schema, the **StructuredOutput call is your entire review — your sole output.** Call it FIRST, before any prose, and do **not** also write the prose report or a prose `## Suppressed` section described below. The orchestrator reads only the structured object; a prose write-up in front of the structured call is wasted work that consumes your turn budget — a reviewer that spends its budget on prose can die before it ever calls StructuredOutput, which surfaces downstream as a missing review. Map your review into the schema: - Each finding → one `findings[]` entry: `severity` (via the mapping under "Severity Levels"), the `Issue` / `Evidence` / `Recommendation` content folded into `description`, and `file` / `line` populated. - Sub-threshold findings → the top-level `suppressed[]` array (one string each), not a prose section. - Grounding citations → the finding's `file` / `line` + `description` (see "Grounding Verdicts" below), not a prose preamble. The prose **"Standard Outp