← ClaudeAtlas

analyze-chapterlisted

Analyzes one manuscript chapter for AI-generation markers across 14 stylistic categories, producing a structured _analysis.md with literal quotes, suggested treatments, and a prioritized verdict. Cross-references the persona, preserve list, and recurrence map before flagging anything. Use when the user asks to analyze a chapter, check a chapter for AI tics, or run analyze-chapter.
rmarsigli/p10t · ★ 0 · AI & Automation · score 62
Install: claude install-skill rmarsigli/p10t
# Skill: analyze-chapter **What it does.** Analyzes one chapter for AI markers and produces a structured `_analysis.md` covering 14 categories, with a verdict and revision priorities. **Triggers** - "analyze chapter X.Y" - "/analyze-chapter X.Y" - "check chapter X.Y for AI tics" - Equivalent phrasing in the project's output language **Input.** Path to the chapter. **Output.** `{chapter}_analysis.md`, at the location set in `.project/config/project.yaml → paths.analyses` — next to the chapter, or centralized in `.project/reports/technical/`. That field is the single source of truth for this choice. > "Next to the chapter" resolves through `paths.layout`: the manuscript root under `flat`, the chapter's own directory under `chapter`. The two fields are **independent axes** and every combination is legal — see `.project/templates/layout.md`. > **Drafting scaffolding is not prose.** `##` scene headers, HTML comment notes, and budget lines are excluded from word counts and from every density figure. Their **absence is never a finding** — a chapter written straight through is a choice, not a defect. See `manuscript/README.md → Scene headers while drafting`. > **Write the analysis in the project's output language** (`config/project.yaml → language`), not in English. --- ## Execution protocol ### Step 1 — Load context Read, in this order: 1. **`.project/config/project.yaml`** — output language, `paths.analyses`, total density ceiling 2. **`.project/templates/framework.md`*