Vibe Coding DiscoverAnti-Autoresearch
View on GitHubDon't trust an autoresearch paper at face value. Reviewer-side integrity forensics (self-consistency + fabrication), deterministic verdict. 61 signals: 46 integrity hack-patterns (families A–H, verdict-bearing) + 13 zero-weight AI writing-style impressions (AIS) + 2 advisory. Not an opaque AI-text classifier. The dual of ARIS.
A Claude Code skill workflow for auditing research papers for self-consistency, fabrication, citation problems, and evaluation flaws. LLM auditors propose evidence-anchored findings; a deterministic adjudicator produces the verdict.
Use Cases
Audit research papers for inconsistencies between claims, tables, and resultsCheck citations for fabricated references or misattributed claimsReview baselines and comparison fairnessInspect research code and results for reproducibility discrepanciesIdentify evaluation design issues such as data leakage or selective reportingGenerate evidence-anchored findings for human paper reviewers
Built With
- Language
- Python
- Frameworks
- Claude Code · MCP · Codex
Tags
agent-skills · research-integrity · paper-review · forensics · Claude Code · peer-review · citation-audit · experiment-audit · deterministic-verdict · AI writing impressions