chapter-lead-writer_skill

Overview

This skill writes concise H2 chapter lead blocks that bind multiple H3 subsections into a coherent argument. It previews the chapter’s comparison lens and 2–3 cross-cutting contrasts without introducing new facts or outline narration. Outputs are body-only markdown files placed immediately after each H2 title.

How this skill works

The skill reads the outline to find H2 chapters with H3s and loads the chapter brief for each (throughline, key contrasts, lead plan). It optionally consults shared writer context for consistent phrasing and validates citation keys against the bibliography. For each target chapter it emits a 2–3 paragraph lead file that sets the lens, connects the H3s, and calibrates evaluation assumptions.

When to use it

• You have H2 chapters whose H3 subsections read like disconnected paragraph islands.
• A chapter brief exists with a throughline and key contrasts (outline/chapter_briefs.jsonl).
• You need a short argumentative bridge that previews contrasts without repeating the TOC.
• You want consistent cross-chapter phrasing using writer context packs.
• You must ensure leads contain no new facts or unsupported citations.

Best practices

• Name 1–2 concrete tensions the chapter resolves and commit to 2–3 cross-cutting contrasts.
• Write an argument bridge that explains why the H3s belong together, not a list of subsections.
• Calibrate comparisons by noting protocol, budget, or tool-access assumptions at a high level.
• Keep leads to 2–3 tight paragraphs and avoid headings or narrative templates.
• Validate any citation keys against citations/ref.bib and only use truly cross-cutting references.

Example use cases

• A methods chapter where H3s cover different evaluation metrics and the lead frames trade-offs between expressivity and verifiability.
• A design chapter whose subsections explore interfaces, planning, and memory—lead ties them via decision-making under budget constraints.
• A results chapter with varied experimental setups—lead clarifies how protocol mismatch and reproducibility shape interpretation.
• A survey chapter that compares architectures—lead previews axes like modularity, interpretability, and operational cost.