ai-wiki-consolidate-drafts

name	ai-wiki-consolidate-drafts
description	Use when handle-local AI wiki drafts have grown into a noisy backlog and related notes should be clustered, refined, or promoted without directly rewriting shared docs by default.

AI Wiki Consolidate Drafts

Use this skill when ai-wiki/people/*/drafts/ has accumulated related notes and the next step is to consolidate them into clearer draft clusters, promotion candidates, or explicit conflicts and supersessions.

Workflow

1. Read the relevant draft set, usually under one handle in ai-wiki/people/*/drafts/.
2. Read nearby context that may confirm or narrow the consolidation target:
   • ai-wiki/conventions/index.md
   • ai-wiki/decisions.md
   • ai-wiki/review-patterns/index.md
   • ai-wiki/problems/index.md
   • ai-wiki/features/index.md
   • relevant files under ai-wiki/trails/ when chronology matters
   • ai-wiki/metrics/ and _toolkit/metrics/*.json only as weak signals
3. Cluster drafts by durable topic, repeated failure mode, or repeated rule, not by filename alone.
4. For each cluster, choose one next action using references/candidate-types.md.
5. Use references/promotion-targets.md to propose the best destination when a cluster is mature enough for shared or repo-level memory.
6. Use references/conflict-and-supersession.md to flag conflicts, refinements, or supersession instead of silently merging away differences.
7. Default to refining handle-local drafts first. Do not directly rewrite ai-wiki/conventions/*.md, ai-wiki/review-patterns/*.md, ai-wiki/problems/*.md, or ai-wiki/features/*.md unless the user explicitly asks for promotion or doc creation.
8. Emit the result using references/output-contract.md.

Rules

• Cluster first, judge second.
• Do not promote every repeated draft.
• Metrics may support prioritization, but they are weak signals and must not be the sole basis for promotion.
• Make the suggested target path explicit whenever the action is Refine draft or Promotion candidate.
• If a cluster is still unstable, keep it local and explain what evidence is still missing.
• Shared conventions and review patterns still require human confirmation before writing shared docs.