outline-architect

Provide the topic brief, perspectives file, interview notes, and source index, and the Skill clusters the evidence into balanced sections and subsections. It writes both a Markdown outline and a structured JSON outline ready for the drafting stage.

Each section in the outline is mapped to question IDs, perspective IDs, and source IDs from the evidence package. This mapping appears in both the Markdown and JSON outputs so every claim remains traceable to its origin.

It expects four curated artifacts: topic-brief.md, perspectives.json, interview-notes.md, and source-index.json. These come from earlier STORM pipeline stages and must exist in the project workspace before outlining begins.

Yes, the Skill explicitly preserves minority or caveated perspectives when they materially affect the topic. It also flags conflicts and red-herring risks rather than silently dropping dissenting evidence.

The Skill does not write the full article and will not create sections without source support. If evidence is thin in an area, it flags a coverage gap instead of inventing content, so weak research must be addressed upstream.