Skip to content

use-cases.md

Latest commit

 

History

History
69 lines (45 loc) · 2.8 KB

use-cases.md

File metadata and controls

69 lines (45 loc) · 2.8 KB

Use Cases

Pre-Submission Manuscript Review

Use pcr before journal submission to check whether tables, manuscript text, raw data, scripts, references, figures, and supplements contain review leads that should be resolved before submission.

Good fit:

  • A manuscript package with source tables, analysis scripts, figures, and references.
  • A lab wants a documented pre-submission checklist before the corresponding author signs off.
  • A researcher wants to know which issues are mechanical consistency problems and which require statistical or domain review.

Not enough by itself:

  • A final accusation, misconduct finding, or publication decision.
  • A review where the source data, manuscript tables, or figure originals cannot be inspected.

Useful commands:

pcr-audit project path/to/project_folder --out build/project.md --json build/project.json
pcr-audit provenance record path/to/project_folder --json build/provenance-record.json

Lab or Institutional Reproducibility Screening

Use project audits and provenance records to review a research package before internal sign-off. Keep all private data local and disable external lookups when required.

pcr-audit project path/to/project_folder --out build/project-offline.md --json build/project-offline.json --no-external-lookups

Editorial or Peer Review Support

Use pcr to produce cautious, evidence-linked review leads. Reports should describe anomalous signals, possible normal explanations, and suggested checks. They should not accuse authors or assert misconduct.

Recommended workflow:

  1. Route the submitted material and record skipped or unsupported checks as info.
  2. Run applicable checks and preserve the JSON output.
  3. Verify high-impact findings against the original source files.
  4. Convert findings into cautious review language using Interpretation boundaries.

Useful inputs:

  • Manuscript DOCX/PDF files.
  • Supplementary CSV/XLSX tables.
  • Statistical text exports.
  • Figure source images.
  • Analysis scripts.

AI-Assisted Audit Orchestration

Agents should read route JSON and run only tools marked ready. Agents can help merge outputs, write cautious summaries, and suggest next review steps, but tool applicability should come from pcr-audit route, not from an agent guess.

pcr-audit route path/to/material --json build/route.json
pcr-audit run path/to/material --scenario auto --out build/audit.md --json build/audit.json

See llms.txt for the concise AI-agent entry point.

Teaching and Benchmarking

The benchmark fixtures provide examples for teaching statistical consistency checks, weak-signal triage, provenance tracking, and responsible interpretation.

python3 benchmark/run_benchmark.py --no-network