← ClaudeAtlas

consolidate-stylelisted

Mines the accumulated decision history — R annotations, revision log, the author's own rewrites — for patterns (three occurrences make a pattern; one makes an anecdote) and proposes evidence-backed updates to persona.md, applied only after the author's ruling. Use when the user asks to consolidate the style, or every 3-5 revised chapters.
rmarsigli/p10t · ★ 0 · DevOps & Infrastructure · score 62
Install: claude install-skill rmarsigli/p10t
# Skill: consolidate-style **What it does.** Periodically reads the accumulated decision history — `R:` annotations, revision log, the author's own rewrites — finds patterns, and proposes evidence-backed updates to `.project/config/persona.md`. The engine behind `define-persona`'s update mode: this skill builds the diff; that one applies it through conversation. **Triggers** - "consolidate the style" - "/consolidate-style" - "what has the system learned about my voice?" - Suggested automatically after every 3–5 chapters revised - Equivalent phrasing in the project's output language **Input.** None required; optionally a scope (since last consolidation — the default). **Output.** A proposed diff to `persona.md`, presented for the author's ruling. Applied only after approval, with a changelog entry. > **Write the proposal in the project's output language.** --- ## Core principle > **Three occurrences make a pattern. One makes an anecdote.** The persona must not chase every decision — a single "kept it" may be situational, a single rewrite may be mood. This skill's value is the **threshold**: it only proposes what the evidence supports, and it shows the evidence. Every proposed change carries its receipts: the decisions, chapters, and quotes that back it. A persona built on receipts is one the author trusts — and one `draft-scene` can bet on. --- ## Execution protocol ### Step 1 — Load context 1. `.project/config/project.yaml` — output language 2. `.project/config/