Repository Public Research preview
1# manuscript-peer-review23An Agent Skill for reviewing a scholarly manuscript (LaTeX or Markdown) with a fixed protocol:4a deterministic audit first, then five independent role reviews, adversarial verification of5every finding, an anchored scoring rubric and an author-facing report.67The protocol exists to prevent the usual failures of machine-assisted review: findings that8quote text which is not in the manuscript, references described without being opened, "wrong"9where "not shown" was meant, and a long undifferentiated list in place of the two or three10comments a decision turns on.1112## Contents1314| Path | Purpose |15|---|---|16| `SKILL.md` | The rules, the eight stages, the finding format, the limits |17| `references/roles.md` | What each of the five roles reads for, and how to delegate them |18| `references/procedure.md` | Each stage in detail, including verification and revision checks |19| `references/rubric.md` | Seven dimensions, anchored levels, weights, caps, how to report scores |20| `references/templates.md` | Intake note, referee report, decision letter, replication checklist, ledger, response table |21| `references/finding-codes.md` | The 67 codes of the audit script and what to check before reporting each |22| `scripts/manuscript.mjs` | Command-line audit tool, JSON or text output |23| `scripts/lib/` | The audit engine (shared with the `manuscript-audit-mcp` server) |2425No dependencies beyond Node.js 20+. No network access, no API key, no LaTeX installation.2627## Install2829```bash30skills install @spboucher/manuscript-peer-review --client claude-code # or codex, cursor, generic-agent-skills31```3233## The script3435```bash36node scripts/manuscript.mjs audit paper.tex --text # every check37node scripts/manuscript.mjs audit paper.tex --root .. --min-severity warning38node scripts/manuscript.mjs citations paper.md --bib refs.bib39node scripts/manuscript.mjs scaffold paper.tex > report.md # report skeleton40node scripts/manuscript.mjs compare v1/paper.tex v2/paper.tex # revision check41node scripts/manuscript.mjs codes # all finding codes42```4344Exit code 0 when no finding reaches the `--fail-on` level (default `error`), 2 when one does,451 on a usage or read error. Files are read only from `--root` (default: the directory of the46manuscript), symbolic links included.4748## Tested4950`node --test skill-tests/` in the source repository: the skill follows the Agent Skills51specification, every referenced file exists, the engine is byte-identical to the MCP package52(whose own suite checks each audit against manuscripts with planted defects), the engine imports53nothing outside Node.js, and the command-line tool is exercised on the same fixtures (output,54exit codes, root confinement).5556The protocol itself (roles, rubric, templates) is a method, not a measured instrument: no57claim is made that its scores are calibrated against editorial outcomes.5859## Limits6061- The script reads LaTeX and Markdown source only: no PDF, no Word files, no images, no data.62- It compiles nothing and looks nothing up. A complete, well-formed reference can still be63 wrong or non-existent; opening the source is the reviewer's job.64- About half of its checks are heuristics on English wording and are labelled as such.65- An empty audit never means a manuscript is correct.6667## Licence6869MIT.70