skills.rest
Light0305
Light0305COMMUNITY
Explore all skills by this author →
GitHub

light-paper-writing

Drafts and validates research papers by binding every claim to evidence and matching wording strength to evidence grade.

572|72|Updated Jun 7, 2026
Search Tags:#peer review#hedging#research integrity#claim evidence#overclaim detection#Academic writing#paper draft
One-click install
npx skills add https://github.com/Light0305/Light-skills --skill light-paper-writing
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: light-paper-writing
Source: https://github.com/Light0305/Light-skills/tree/main/skills/light-paper-writing
Command: npx skills add https://github.com/Light0305/Light-skills --skill light-paper-writing

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Academic drafts often contain strong claims with no supporting evidence, wording that overstates weak results, and contributions that drift between abstract, introduction, and conclusion. This Skill turns those integrity risks into deterministic machine-readable gates so a manuscript can be checked before submission.

Core Features & Use Cases

  • Claim-evidence gate: Binds each strong assertion in a draft to its own evidence IDs via a light.paper_claims.v1 claim map with draft SHA-256, flagging unsupported claims as critical and overclaiming as warnings.
  • Wording and consistency linting: Checks that hedging matches evidence strength (strong/moderate/weak/none), detects contribution drift across abstract/introduction/conclusion, and runs offline mechanical checks for AI-style phrasing, passive voice, and spin.
  • Reviewer-perspective workflow: Provides IMRaD and conference templates, argument-outline contracts, self-review checklists, and rerouting back to result-analysis or experiment stages when evidence is missing.
  • Use Case: After finishing experiments, you write a draft claiming a 3.1-point improvement. The gate verifies the claim binds to a real evidence entry and result card, warns if the conclusion weakens the number, and blocks finalization if any strong assertion lacks evidence.

Quick Start

Ask the AI to draft or review your paper with this skill, providing your draft markdown and the evidence_strength.json from result analysis so every claim is checked against its evidence.

Frequently Asked Questions about light-paper-writing

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check that every claim in my paper draft has evidence?▼

Build a light.paper_claims.v1 claim map linking each draft sentence to evidence IDs from evidence_strength.json, then run claim_evidence_gate.py with your draft and claim map. Strong assertions without bound evidence are reported as critical findings.

How to match wording strength to statistical evidence in academic writing?▼

The gate compares each claim's wording against its evidence grade: strong evidence allows demonstrate, weak evidence requires hedges like suggest, and non-significant results may only report no significant difference. Overstated wording is flagged as an overclaim warning.

Does the paper writing gate work offline without API keys?▼

Yes, mechanical_check.py and polish.py run fully offline using only the Python standard library. The optional LanguageTool online mode is opt-in via a flag and rate-limited for single documents.

Can it detect contribution drift between abstract and conclusion?▼

Yes, contribution_consistency.py extracts contribution statements from abstract, introduction, and conclusion and reports number drift, strength drift, and coverage drift as warnings. Final judgment on alignment still requires human review.

What are the limitations of automated claim-evidence checking?▼

The gate verifies that evidence bindings exist and wording matches the declared grade, but it does not verify whether the evidence itself is correct or the logic is sound. Judgments about novelty, argumentation quality, and persuasiveness still require human reviewers.

Related Skills

Explore More Education & Research Skills→
EvoClawEvoClaw
11

claim-evidence-alignment

Validate manuscript claims against linked figures, tables, experiments, and statistical results.

Official
Advanced
jakechen1jakechen1

paper-orchestra

Transform research ideas and experimental data into submission-ready LaTeX manuscripts.

Community
Advanced
Tnemo65Tnemo65

academic-pipeline

Coordinate end-to-end academic research workflows from discovery to manuscript with integrity checks.

Community
Advanced
wen-chenwen-chen
7

dr-cook:english-polisher

Polish academic English grammar, word choice, and sentence structure across research domains.

Community
Advanced