- Home
- Skills
- Willoscar
- Research Units Pipeline Skills
- Style Harmonizer
style-harmonizer_skill
- 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 style-harmonizer- SKILL.md7.3 KB
Overview
This skill removes generator-style slots and repetitive discourse stems across flagged sections/*.md while preserving meaning and all citation keys. It performs targeted surface rewrites to harmonize cadence, openers, and connectors without adding facts or moving citations. Use it when multiple subsections share a mechanical rhythm that undermines reader trust.
How this skill works
The skill reads output/WRITER_SELFLOOP_TODO.md to find the Style Smells list, then edits only the referenced sections/*.md files. Edits are limited to sentence-level reshaping: replace count-based openers, vary discourse stems, and alter opener cadence while keeping claim→evidence anchoring intact. It leaves citation keys and factual content untouched and writes a sections/style_harmonized.refined.ok marker when done.
When to use it
- writer-selfloop is PASS but output/WRITER_SELFLOOP_TODO.md flags Style Smells
- Draft sections read like they share the same opener/cadence across H3s
- You need surface harmonization without new evidence or citation changes
- You want to reduce generator-voice signals (count-based openers, reused stems)
Best practices
- Only edit files named in the Style Smells section; do not touch other files.
- Make small, local edits: change opener shapes, move connectors into clauses, vary sentence cadence.
- Never add, remove, or relocate citation keys; spot-check cited claims after each edit.
- If you must introduce new content to fix clarity, stop and route upstream for evidence (C3/C4).
- Prefer content-bearing pivots (tension-first, decision-first, protocol-first) instead of generic narrations.
Example use cases
- Replace repeated "Two limitations... First,..." openers with a single caveat integrated into the paragraph.
- Turn repeated "The key point is that..." stems into varied takeaways like "A practical implication is..." or split claim + why sentences.
- Swap paragraph-leading adverb connectors with subject-first sentences that embed the relation mid-sentence.
- Convert overview/toC narration openers into a falsifiable lens that states a constraint or decision affecting interpretation.
- Rename internal shorthand like "tokens" into reader-facing nouns (protocol details, parameters) when context permits.
FAQ
No. The skill preserves all citation keys and does not move citations across sentences or subsections.
What if harmonizing requires new evidence?
Stop and route upstream. This skill does not add facts or citations; if additional evidence is needed, escalate to C3/C4.