reviewlisted

# design-doc-review:review Coordinator for the design-doc-review pipeline. Runs each check as an independent lean agent (see the `lean-agent-loop` skill — read it if you haven't loaded it this session), collects small structured summaries, then drives a fix→reverify loop until every check passes or a round cap is hit. **Target**: {{args}} — a file path to the design doc. If omitted, use the doc already in context. This skill does not invent checks — it dispatches to whatever lives under `design-doc-review/skills/*` other than itself. Today that's `outline`, `readability`, and `visuals`. Adding a new check later (e.g. `evidence-quality`, `risk-completeness`) means adding one `skills/<name>/SKILL.md` following the same JSON-summary contract — no change needed here beyond adding it to the registry below. ## Registry | Check | Skill | Fixable automatically? | |---|---|---| | Structure / topic coverage | `design-doc-review:outline` | Rarely — most gaps need author input (e.g. "what's the actual rollback plan"). Coordinator proposes a stub + a question, does not invent content. | | Prose / cognitive load | `design-doc-review:readability` | Yes — filler removal, hedge cleanup, front-loading, splitting mixed-purpose paragraphs are mechanical rewrites. | | Missing diagrams / comparison tables | `design-doc-review:visuals` | Rarely — coordinator can propose a mermaid/table skeleton with extracted axes, but correct diagram type and exact table columns need author confirmation; neve