Paper Claim Audit
Checks every number and comparison in a paper against raw result files using a fresh reviewer.
- Stars
- 18,482
- Licence
- MIT
- Files
- 1 file
What it does
Once installed, the agent can check a research paper draft against its raw experiment output files. It pulls out each number, comparison and scope claim (for example rounding errors, best-seed cherry-picking, wrong deltas, or 'consistently outperforms' on only two datasets) and traces it to the evidence. To avoid bias, it hands the job to a separate reviewer model that sees only the .tex files and raw results, not the executor's notes or summaries.
Great for
- Researchers preparing a paper for submission who want a second check that numbers match their data
- Teams using an automated research pipeline who want the paper-writing step verified independently of the experiment step
- Students checking that their reported results match their logs and result files
Skip it if
- Judging whether the experiment design or scientific conclusion is sound; the skill only checks paper-to-evidence fidelity (use the separate experiment or result-to-claim audits for that)
- Papers with no raw result files in the repository, since there is nothing to trace claims to
- Running on a timer or in a loop; the skill itself says it should be run once when the draft changes, not repeatedly
Try asking
- “Check paper claims before I submit to the conference”
- “Verify numbers in paper/main.tex against the results folder”
- “审查论文数据”
- “Audit whether the 'improves by 15%' claim in the abstract matches the result files”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
paper-claim-audit.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip paper-claim-audit.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip paper-claim-audit.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Run it once the paper draft and results are final, not on a schedule, because the verdict only changes when the paper or data changes.
- Keep executor summaries such as EXPERIMENT_LOG.md out of the results folder so the reviewer stays zero-context.
- Check the reviewer backend and model are set up as you expect, since the skill sends paper and result content to an external reviewer tool.
What's inside
- SKILL.md
Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.