| name | deviation-impact-analysis |
| description | Turn "AI draft -> author edits" diffs into structured deviation items, an author-intent prompt, and a concrete propagation plan. |
When to use
- After author edits a chapter draft (especially via external editor like Cursor)
- Before auto-continue to future chapters
Outputs
deviation_report.md(human readable summary)deviation_prompt.md(one-click prompt/instructions)change_impact_report.md(what future outline/chapters/canon/ledger will be affected)backup_options_report.md(conservative vs aggressive vs branch)
Structured deviation categories (must classify)
- plot / character / tone / info density / canon setting / timeline / style / pacing
Procedure
- Extract diff signals and cluster them into 5-20 deviation items (avoid noise).
- Summarize author intent in a short instruction that can drive future rewriting.
- Scan future outline/chapters/canon/ledger:
- list concrete touch points (file/section) that need change
- Produce 2-3 backup routes and recommend one.