ldm2060/research_copilot/de-ai-checker
Claude Code and agent skill by ldm2060/research_copilot
- community
Checking that the SKILL.md file is still in the repository.
Validation skill for de-AI quality checking. Use AFTER polishing to verify that AI patterns have been removed and text reads naturally. Triggers on: 'check de-AI quality', 'verify humanization', 'validate polish', 'de-AI checker', '检查去AI效果'. This is a validation-gate compatible wrapper around paper-deai.
Install
npx vskill@latest install ldm2060/research_copilot/de-ai-checkerSKILL.md
The same SKILL.md was published earlier as ldm2060/research_copilot/de-ai-checker.
De-AI Checker — Post-polish validation
Run after polishing is complete to verify that AI patterns have been successfully removed and the text reads naturally.
When this skill fires
Fire automatically when:
- The polisher agent reaches the VERIFYING state and needs validation-gate clearance
- User explicitly requests de-AI quality verification
- After any
paper-deairewriting pass to confirm quality
Also fires on user request: "检查去AI效果" / "check de-AI quality" / "verify the polish worked"
Purpose
This skill serves as a validation-gate compatible checker that:
- Verifies AI patterns have been removed
- Confirms text naturalness and academic register
- Provides pass/fail validation for capability gates
- Delegates actual rewriting to
paper-deaiif issues are found
Procedure
Step 1 — Read the polished text
Load the text that was just polished. This should be LaTeX content from the target sections.
Step 2 — Run AI pattern detection
Check for common AI fingerprints:
| Pattern Category | What to check | |
|---|---|---|
| Over-used vocabulary | Scan for the AI-over-used word list: leverage, delve, tapestry, endeavor, underscore, etc. (full list in paper-deai skill) | |
| Mechanical connectives | Look for: "First and foremost", "It is worth noting that", "Moreover", "Furthermore" at sentence starts | |
| List format | Check for \item usage where prose would be more natural | |
| Excessive dashes | Count em-dashes (--- and —) — more than 1 per paragraph is suspicious | |
| Tense errors | Verify background/prior-art uses present perfect ("have achieved") not simple present | |
| Emphasis abuse | Check for bold/italics used for emphasis rather than structure |
Step 3 — Score and decide
Calculate a naturalness score based on:
- AI vocabulary count (each instance: -10 points)
- Mechanical connectives (each instance: -5 points)
- List format in prose sections (each
\item: -8 points) - Excessive dashes (each beyond 1/paragraph: -3 points)
- Tense errors (each instance: -7 points)
Pass threshold: Score ≥ 85/100
Step 4 — Output validation report
Step 5 — Auto-fix if requested
If the validation fails and the user/agent requests auto-fix:
- Call the
paper-deaiskill with the problematic sections - Re-run this validation checker
- Report final status
Output format
- Part 1 [Validation Report]: The structured report above
- Part 2 [Gate Status]: One of:
[VALIDATION_GATE_PASS]— text is natural, no AI patterns detected[VALIDATION_GATE_FAIL]— AI patterns remain, re-polish required
Integration with capability gates
This skill satisfies the validation-gate requirement because:
- Name matches
*-checkerpattern - Provides explicit pass/fail status
- Can be called via
Skill(skill='de-ai-checker')
Hard constraints
- Post-polish only — if no polished text exists, exit and recommend running
paper-deaifirst - Read before checking — every issue must cite file:line, not "in general"
- Honest scoring — do not inflate scores to declare victory
- Delegate rewriting — this skill validates;
paper-deairewrites - One pass per invocation — do not loop; if re-check is needed, user/agent re-invokes
- Explicit gate status — always output
[VALIDATION_GATE_PASS]or[VALIDATION_GATE_FAIL]
Relationship to paper-deai
paper-deai: Executor — rewrites text to remove AI patternsde-ai-checker: Validator — checks quality and provides gate clearance
Use paper-deai for rewriting, use de-ai-checker for validation.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingself/skills/de-ai-checker/SKILL.md
Popularity
- 7-day trend score
- 0
Refreshed 15 Nov 2026, 3 days ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/ldm2060/research_copilot/de-ai-checker)