Skip to main content
skill CX Codex

ccf-integrity-auditor — Claude Code skill: audit claims

ccf-integrity-auditor is a Claude Code skill that checks claims, figures, terms, and citations against supplied evidence. It audits manuscripts and research results.

MIT 2,200 installs ★ 3,769 updated 2026-10-20

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add mikubaka88/CCFA-Skills --skill ccf-integrity-auditor -a claude-code
Open on GitHub

What ccf-integrity-auditor does

ccf-integrity-auditor checks whether manuscript claims are supported by reported results and whether citations match real works and their cited context. It compares values, units, metric direction, tables, figures, captions, abstracts, and conclusions, separating inconsistent figures from values that cannot be compared.

The skill has claim-audit, numeric-audit, citation-audit, and full modes. Citation checks cover BibTeX metadata, duplicate keys, DOI, arXiv, venue details, and support from primary sources. A run can use a manuscript, figures and tables, results, a bibliography, ccfa.yaml, and the requested audit mode.

Findings include source locations, affected claims or values, evidence, severity, and unverified scope. The audit does not authorize manuscript rewriting; edits require separate authorization.

What it helps with

  • trace manuscript claims back to experimental evidence
  • reconcile values across prose, tables, and figures
  • check whether existing citations support their surrounding claims
  • spot BibTeX metadata errors and duplicate keys
  • separate unsupported claims from overstated ones

Example prompts for the agent

  • Audit the claims in paper.tex against the results in results.csv, and mark unsupported statements.
  • Run numeric-audit across tables/, figures/, and the manuscript text to compare values, units, and metric direction.
  • Use citation-audit on references.bib and check each citation's context against primary sources.

How to install ccf-integrity-auditor

1Skills CLI (Vercel)

npx skills add mikubaka88/CCFA-Skills --skill ccf-integrity-auditor -a claude-code

2Manually: put the files in place

~/.claude/skills/ccf-integrity-auditor/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/ccf-integrity-auditor/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about ccf-integrity-auditor

How do I install ccf-integrity-auditor in Claude Code?

Run npx skills add mikubaka88/CCFA-Skills --skill ccf-integrity-auditor -a claude-code, then check ~/.claude/skills/ccf-integrity-auditor/SKILL.md.

What does ccf-integrity-auditor do?

The skill maps important claims to evidence, reported numbers to results, and existing citations to real papers and their context. It also checks BibTeX details and consistency across manuscript sections, tables, and figures.

What does ccf-integrity-auditor need to work?

It needs a manuscript, related figures and tables, results, a bibliography, ccfa.yaml, and an audit mode. Citation-context checks require the relevant primary sources.

Source text

SKILL.md · MIT
---
name: ccf-integrity-auditor
description: "Audit existing CCF claims, numbers, terminology, and citations against supplied or verified evidence. Use for 引用核验, claim审计, 数字一致性, and BibTeX/context checks. Full scientific review belongs to ccf-paper-reviewer; new literature discovery belongs to ccf-literature-searcher."
metadata:
  ccf_skill_controls:
    handoff_question_mode: partial
    respect_session_denylists: true
    protect_idea_scope_in_writing: true
    private_material_safety: moderate
    shared_controls: ../ccf-common/references/
---

# CCF Integrity Auditor

## Family File Contract

Before file work, reuse canonical paths and one task/artifact working directory; default to project-root `ccfa-workfiles/<purpose>/<artifact-id>/`. Update in place, preserve inputs, and use UTF-8 with Chinese-text checks. Apply [artifact-contracts.md](../ccf-common/references/artifact-contracts.md) for naming, ownership, and task-created cleanup; handoffs inherit these paths.

## Collaboration Contract

Before specialist execution, read and apply [ccf-humanization](../ccf-humanization/SKILL.md) first, then [ccf-common](../ccf-common/SKILL.md). At every handoff, reuse their applicable active rules or refresh missing/changed ones. Both preflights are required even without prose; detailed editing, experiment, and maintenance modes run only when relevant.

Integrate necessary specialist contributions under one owner; reuse valid prerequisites and verify affected results. Follow the [cooperation routes](../ccf-common/references/routing.md) without duplicate reports.

## Core Rule

Trace each important claim to supplied evidence, each number to supplied results, and each citation to a real cited work and a supported citation context. Mark unsupported items instead of repairing them by invention.

## Modes

- `claim-audit`: claim-support and result-to-claim consistency.
- `numeric-audit`: numbers, units, table/figure/text agreement, deltas, and metric direction.
- `citation-audit`: already cited papers, BibTeX metadata, duplicate keys, DOI/arXiv/venue sanity, and citation-context support.
- `full`: all integrity checks.

## Workflow

1. Identify supplied manuscript, figures/tables/results, bibliography, `ccfa.yaml`, and requested audit mode.
2. For claim or full audits, map the relevant claims to evidence and mark each as supported, partially supported, unsupported, overstated, or unclear. Numeric and citation audits track their scoped values or citations; add claim rows only when needed to resolve a finding.
3. Cross-check the reported values within the requested scope across text, tables, figures, captions, abstracts, and conclusions. Use deterministic arithmetic for deltas, units, metric direction, and rounding. Distinguish not comparable from inconsistent.
4. For citation audit, verify the identity and context support of existing citations through primary sources; metadata existence alone does not establish support. Batch independent identifiers when possible and request only identity fields before reading the source passages needed for claim support. Reuse verified source versions and anchors; refresh affected findings when a citation, claim, or source version changes. Seek new literature only when requested; broad discovery belongs to `ccf-literature-searcher`.
5. For any questionable citation, separate metadata problems from context-support problems.
6. Hand off to `ccf-paper-reviewer` for full scientific judgment and to `ccf-paper-writer` for safe wording edits.
7. If the numbers and claims are consistent but the figure/table layout, caption placement, palette, float order, or rendered readability is weak, hand off to `ccf-visual-composer`.

## Output Contract

Include the fields relevant to the requested audit, with source locations and unresolved coverage. A bounded helper returns its findings to the owner without a separate full report.

```text
Mode:
Artifacts checked:
Claim-evidence matrix:
Numeric consistency findings:
Citation metadata findings:
Citation-context findings:
Severity:
Safe edit suggestions:
Next CCFA owner:
Unverified scope:
```

## Execution Boundaries

Follow `../ccf-common/references/handoff-modes.md`, `../ccf-common/references/task-modes.md`, and `../ccf-common/references/privacy-and-evidence.md`. Finish the checkable portions when an attachment or source is missing. Report exact file/page/table locations, affected values or claims, evidence, and severity; mark unverified coverage separately from failures. An audit request does not authorize manuscript rewriting. Use existing authorization for explicitly requested fixes and preserve raw measurements.

© mikubaka88 · MIT · mikubaka88/CCFA-Skills

Similar skills

All skills for Science and research →

From the same set18

The whole set →
ccf-common ccf-common sets baseline preflight checks, routing, and artifact contracts for CCFA skills. Runs right after ccf-humanization before handing tasks to research specialists. CX Codex +1 2,300installs ccf-submission-checker ccf-submission-checker — a Claude Code skill checks CCF venue requirements and submission-package readiness. Run it to review formatting, a TeX/PDF build, or artifact reproducibility. CX Codex +1 2,200installs ccf-rebuttal-writer ccf-rebuttal-writer — a skill for automating review rebuttals and revision tracking in academic submissions. It generates response letters, resubmission plans, and revision ledgers while adhering to formatting standards. CX Codex +1 2,200installs ccf-pipeline-orchestrator The ccf-pipeline-orchestrator Claude Code skill plans research stages and coordinates specialist skills. It handles task breakdown, project tracking, and explicitly requested autonomous research. CX Codex +1 2,200installs ccf-literature-monitor The ccf-literature-monitor Claude Code skill tracks recent research and checks papers for overlap with an idea. It scans sources such as arXiv, OpenReview, conference pages, labs, and named competitors. CX Codex +1 2,200installs ccf-visual-composer ccf-visual-composer renders Python data analyses and generates scientific diagrams through a three-stage GPT Image workflow. Triggered when building architecture diagrams or redrawing figures. CX Codex +1 2,200installs ccf-idea-reviewer The ccf-idea-reviewer Claude Code skill assesses research ideas for value, novelty, insight, and mechanism. It reviews concepts by default, without judging experiments. CX Codex +1 2,200installs ccf-skill-forger ccf-skill-forger — a skill for Claude Code, Codex, Cursor and other agents that audits and maintains the CCFA skill family: structure, triggers, dependencies, and SVG diagrams. CX Codex +1 2,200installs

Project sponsors

Companies that keep this analytics open