← ClaudeAtlas

review-revisionlisted

Evaluates the author's revision of a chapter against the R-annotated analysis file — introduced errors, rewrite quality, inverted problems, residual density, continuity gaps — and explicitly answers every question the author left in their annotations. Use when the user asks to check their notes and a chapter, review their revision, or evaluate their edits.
rmarsigli/p10t · ★ 0 · Code & Development · score 62
Install: claude install-skill rmarsigli/p10t
# Skill: review-revision **What it does.** After the author revises a chapter using an `_analysis.md` annotated with `R:`, this evaluates the result: what works, what needs adjustment, errors introduced during revision, and whether the quality target was met. **Triggers** - "check my notes and chapter X" - "review my revision of X.Y" - "/review-revision X.Y" - "how did chapter X turn out after my edits" - Equivalent phrasing in the project's output language **Inputs** - Path to the revised chapter - (Implicit) The matching `_analysis.md` with `R:` annotations **Output** - Structured chat response - **An entry in `.project/reports/revision-log.md` — mandatory, not optional.** See "Revision log entry" below. > **Write all output in the project's output language.** --- ## Why this skill exists The cycle is: 1. `analyze-chapter` produces the report 2. The author decides item by item, annotates `R:` 3. The author rewrites 4. **`review-revision`** closes the loop Without step 4, the author revises blind. This skill is the **fresh pair of eyes** that says "this landed, this got worse, and you left an agreement error here". It is also where the system **learns**: author decisions feed `persona.md` and `preserve-list.md`. --- ## Execution protocol ### Step 1 — Load context 1. `.project/config/project.yaml` — output language, total density ceiling 2. The `_analysis.md` with `R:` annotations — including its recorded word count and per-category densities, which are the "be