outline-refiner_skill

This skill provides an auditable planner pass for outlines by reporting coverage, reuse hotspots, and axis specificity without prose.
  • Python

109

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill willoscar/research-units-pipeline-skills --skill outline-refiner

  • SKILL.md3.5 KB

Overview

This skill runs a deterministic planner pass that audits an outline and its mapping, producing a coverage report and a run-state record. It flags low evidence per subsection, reuse hotspots, scaffold-y axes, and structural mismatch with a target ToC budget. The outputs are machine-verifiable artifacts, not prose.

How this skill works

It parses outline/outline.yml to enumerate H2/H3 structure and uses outline/mapping.tsv to compute per-H3 coverage and paper reuse statistics. Optional inputs (paper notes, subsection briefs, budget report, GOAL.md) are ingested to add evidence-level, axis-specificity, and change-log context. The skill writes outline/coverage_report.md (bullet diagnostics and small tables only) and appends a JSONL record to outline/outline_state.jsonl.

When to use it

  • After section-mapper produced outline/outline.yml and outline/mapping.tsv
  • Before starting draft writing to confirm mapped evidence is sufficient
  • When you need a verifiable, NO-PROSE planner pass that guides structural edits
  • When you suspect repeated reuse of a small set of papers across subsections
  • When you want an auditable run history of outline refinement

Best practices

  • Provide optional inputs when available: papers/paper_notes.jsonl and outline/subsection_briefs.jsonl to enable richer diagnostics
  • Keep the final ToC target in mind (≈6–8 H2); use the report to merge or split sections accordingly
  • Treat the report as prescriptive bullets only—do not expect narrative prose or invented citations
  • Respect the freeze policy: if outline/coverage_report.refined.ok exists, rerun only after intentional unfreeze
  • Run the skill iteratively after mapping updates to track coverage changes over time

Example use cases

  • Detect H3s with zero or single-paper coverage and list actions (map more evidence, merge, or drop)
  • Identify reuse hotspots where the same papers are mapped to many H3s and suggest diversification actions
  • Flag generic subsection axes (scaffold-y) and propose axis-specific rewrites as bullet items
  • Compare current structure to OUTLINE_BUDGET_REPORT.md to explain recent merges and recommend further consolidation
  • Produce an append-only outline/outline_state.jsonl entry for CI auditing and reproducibility

FAQ

No. The planner pass produces bullet diagnostics and small tables only. It does not write survey prose or invent papers.

What triggers a non-overwrite of the coverage report?

If outline/coverage_report.refined.ok exists, the script will not overwrite outline/coverage_report.md; this enforces freeze policy.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
outline-refiner skill by willoscar/research-units-pipeline-skills | VeilStrat