outline-budgeter_skill

Overview

This skill rewrites an over-fragmented outline into a paper-like section budget without producing prose. It reduces H2/H3 explosion by merging adjacent or weak subsections to produce a target Table of Contents (~6–8 H2, fewer thicker H3) while preserving scope and evidence needs. It emits a short bullets-only report documenting merges and risks.

How this skill works

The skill reads outline/outline.yml and optional signals (queries.md for draft_profile, outline/mapping.tsv, outline/coverage_report.md, GOAL.md) to compute a budget snapshot: H2 count, total H3 count, and H3 per H2. It plans evidence-aware merges (prefer adjacent subsections, weakly-covered items, or high-reuse fragments) and rewrites the outline in place, consolidating titles and bullets and keeping ids stable when possible. It also writes outline/OUTLINE_BUDGET_REPORT.md listing before/after counts, merges performed, and any mapping risks.

When to use it

• You have outline/outline.yml that would produce many thin H2/H3 sections
• Before or immediately after running section-mapper to avoid mapping thin subsections
• When queries.md sets a draft_profile (survey or deep) and you need an H3 budget
• If outline/coverage_report.md indicates weak coverage or high-reuse subsections
• When GOAL.md defines scope and you must avoid scope-mixing merges

Best practices

• Prefer merging adjacent H3s that share keywords or comparison axes rather than distant topics
• Use mapping.tsv and coverage_report.md to prioritize merges that reduce weak-covered fragments
• Keep scope constraints from GOAL.md central: do not merge across distinct research questions
• When merging, move fine-grained distinctions into bullets or explicit comparison axes instead of creating new subsections
• If you change section ids, record them in the report and plan to reset mapping (rerun section-mapper)

Example use cases

• Survey outline with 15+ H2s and dozens of one-line H3s; compress to ~6 H2 with thicker H3s for a survey draft_profile
• Deep literature review where queries.md sets deep: lower H3 count to target thicker subsections aligned with evaluation axes
• Outline-refiner output shows many low-coverage H3s; merge weak items into thicker subsections guided by mapping.tsv
• Preparing a draft skeleton before automated prose generation to ensure sections are substantial and evidence-first