agentleFS
wangyi0403wangyi0403/dl-research-workflow/.agents/skills/claim-numerics-audit/SKILL.md

claim-numerics-audit

Skill4 starsChanged 41 days agoMITView on GitHub

Check manuscript numbers against experiment records, figures, tables, configurations, and units; report mismatches without silent edits.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read AGENTS.md, the LaTeX/Markdown manuscript, docs/05_experiment_plan.md, docs/06_result_tables.md, docs/07_experiment_tracker.md, …

    line 10
  2. Distinctive:

    Check value, rounding, unit, denominator, split, seed aggregation, uncertainty interval, and comparison direction.

    line 12
  3. Distinctive:

    Verify superlatives and deltas against the appropriate comparison set.

    line 13
  4. Very specific:

    Write every verified item, mismatch, and unresolved item into docs/11_pre_submission_audit.md under “数字与论断核验”.

    line 14

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Claim Numerics Audit

Audit numerical assertions in the manuscript against their traceable research artifacts. The audit distinguishes results, uncertainty, configuration values, and reference noise instead of matching raw numbers mechanically.

  1. Read AGENTS.md, the LaTeX/Markdown manuscript, docs/05_experiment_plan.md, docs/06_result_tables.md, docs/07_experiment_tracker.md, docs/08_analysis.md, results/data, and the cited figure/table sources.
  2. For each reported numerical claim, record manuscript location, value and unit, comparison direction, Claim ID, and its supporting Experiment ID/table/figure/configuration. Ignore unambiguous equation labels, page numbers, and bibliography years unless they are used as a claim.
  3. Check value, rounding, unit, denominator, split, seed aggregation, uncertainty interval, and comparison direction. Treat a real mismatch, stale source, or untraceable number as an issue; do not set arbitrary tolerances that hide a material discrepancy.
  4. Verify superlatives and deltas against the appropriate comparison set. A task result may support an engineering claim without supporting a mechanism or generalization claim.
  5. Write every verified item, mismatch, and unresolved item into docs/11_pre_submission_audit.md under “数字与论断核验”. Include the resolution owner; never silently edit source code, tables, or manuscript text.

Gate outcome

  • Pass only when all material reported values and directions are traceable and no unresolved discrepancy changes the stated claim.
  • Otherwise report conditional or fail, identify the smallest repair, and preserve the original record.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.