Project Integrity Check

VerifiedSafe

Run a full project integrity check: pipeline and manuscript audit. Report findings without fixing anything without approval.

Sby Skills Guide Bot
Data & AIIntermediate
1608/14/2026
Claude CodeCursorWindsurfCopilotCodex
#project-integrity#pipeline-audit#manuscript-check#reproducibility#file-validation

Recommended for

Our review

This skill performs a two-phase integrity audit of a research project, verifying pipeline files and manuscript references.

Strengths

  • Comprehensive coverage of scripts, data, parameters, and outputs.
  • Cross-checks between the README, manuscript, and actual files.
  • Detects orphan files and undocumented references.
  • No changes are made without user approval.

Limitations

  • Requires an up-to-date and accurate README and table/figure map.
  • May produce false positives if files are intentionally unreferenced.
  • Depends on the quality of existing documentation.
When to use it

Before submission or publication to ensure the project is complete and consistent.

When not to use it

When the project is in active development and files are expected to change frequently.

Security analysis

Safe
Quality score85/100

The skill only describes a manual/computational integrity check process with no destructive or exfiltrating actions. It does not instruct executing shell commands, downloading remote content, or modifying files; it merely asks the agent to verify existence and consistency.

No concerns found

Examples

Full project check
Run a full project integrity check on this repository.
Manuscript reference audit
Check that all scripts and data files from the README pipeline table exist and that every \includegraphics and \input in manuscript.tex resolves to an existing file.
Orphan and undocumented outputs
Audit this project for orphan output files not referenced in the manuscript and any manuscript references missing from the table/figure map.

Run a full project integrity check. Two phases:

Phase 1 — Pipeline integrity: (1) Every script listed in README pipeline table exists. (2) Every input data file listed exists. (3) No scripts or data files sitting in wrong directories. (4) Output directories exist. (5) params.do values match README parameters table.

Phase 2 — Manuscript audit: (1) Every entry in README table/figure map has an existing output file and source script. (2) Every \includegraphics and \input in manuscript.tex resolves to an existing file. (3) Find any output files not referenced in the manuscript (orphans). (4) Find any manuscript references not in the table/figure map (undocumented).

Report all findings. Do not fix anything without user approval.

Related skills