Agent Skills

ccf-integrity-auditor

researchmikubaka881.4K installs

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.

Install

npx skills add https://github.com/mikubaka88/ccfa-skills --skill ccf-integrity-auditor
SKILL.md

CCF Integrity Auditor

Family File Contract

Before writing, resolve the canonical output and one stable working directory per task/artifact. Reuse explicit or established task paths; otherwise use project-root ccfa-workfiles/<purpose>/<artifact-id>/, with source/, assets/, cache/, and build/ only as needed. Update current files in place; do not scatter intermediates or create iteration copies. Preserve inputs and required evidence; clean only verified disposable files created by this task. Use UTF-8 text I/O and check Chinese text after saving or rendering. For file work, apply artifact-contracts.md and reuse the same paths across skill transitions.

Collaboration Contract

Before specialist execution, read and apply ccf-humanization first, then ccf-common. 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.

Keep one integrating owner and actively use other skills to resolve missing prerequisites or check material findings. Reuse applicable evidence; do not skip necessary groundwork to save tokens. Before finalizing, integrate contributions and verify affected results. Follow the conditional cooperation routes; avoid unrelated stages and 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. Build a claim-evidence matrix and mark each claim as supported, partially supported, unsupported, overstated, or unclear.
  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. 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

Mode:
Artifacts checked:
Claim-evidence matrix:
Numeric consistency findings:
Citation metadata findings:
Citation-context findings:
Severity:
Safe edit suggestions:
Next CCFA owner:
No-invention status:

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.

Related skills

researchmattpocock575KInvestigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.paper-context-resolverlllllllama451KRigor Paper Context helper for README-first deep learning repo reproduction. Use only when the README and repository files leave a narrow reproduction-critical gap and the task is to resolve a specific paper detail such as dataset split, preprocessing, evaluation protocol, checkpoint mapping, or runtime assumption from primary paper sources while recording conflicts. Do not use for general paper summary, repo scanning, environment setup, command execution, title-only paper lookup, or replacing Renv-and-assets-bootstraplllllllama450KRigor Setup skill for README-first deep learning repo reproduction. Use when the task is specifically to prepare a conservative conda-first environment, checkpoint and dataset path assumptions, cache location hints, and setup notes before any run on a README-documented repository. Do not use for repo scanning, full orchestration, paper interpretation, final run reporting, or generic environment setup that is not tied to a specific reproduction target.ai-research-explorelllllllama311KRigor Explore compatible skill slug for meaningful and potentially novel deep learning research candidates. Use when the researcher has chosen the task family, dataset, benchmark, evaluation method, provided SOTA references, and wants candidate-only exploration on top of `current_research` with auditable repo understanding, idea gating, fair comparison, and governed experiments written to `explore_outputs/`. Do not use for README-first trusted reproduction, open-ended direction finding, narrow c

Search skills and MCP servers

Fuzzy search across 23,137 skills and servers