| name | review |
| description | Audit a Quarto manuscript or a single notebook — freeze freshness, data-path existence, citation integrity, figure/table export presence, placeholder and anonymization checks. Produces a scored review report under notes/. Read-only — never modifies notebooks, data, or the manuscript. |
| argument-hint | [scope] — 'manuscript' (default, full pre-submission audit) or '<notebook-slug>' (single notebook, e.g. notebook-02) |
| disable-model-invocation | true |
| user-invocable | true |
| allowed-tools | Bash, Read, Glob, Grep, Write |
| version | 1.0.0 |
| workflow_stage | quality |
| tags | ["quality-check","audit","submission","freshness","citations","anonymization"] |
Review: scored audit of the manuscript or a notebook
A single thorough, read-only audit that produces a scored report with a
verdict, a dimension table, an issues table, and priority action items — modeled
on example_project2025b's review skills. Absorbs the former data-audit,
freeze-check, and submission-prep skills.
What this skill does:
- Audits the manuscript (default) or one notebook across the dimensions below.
- Produces a verdict: SUBMISSION-READY / MINOR REVISION / MAJOR REVISION.
- Saves the report to the
notes/ audit trail (see notes/README.md).
What this skill does NOT do:
- Never modifies notebooks,
index.qmd, data, or config. The only file it
writes is the review report. Honor CLAUDE.md (no deletes, no moves).
Example invocations
/project:review # full manuscript pre-submission audit
/project:review manuscript # same as above
/project:review notebook-02 # audit a single notebook
Phase 1 — Pre-flight
- Parse the scope:
manuscript (or empty) → full audit; otherwise treat the arg
as a notebook slug and confirm notebooks/<slug>.qmd exists (stop if not).
- Read
references/audit-checklist.md (procedure for every dimension),
references/scoring-and-criteria.md (severity + verdict logic), and
references/submission-checklist.md (manual human-action items).
- Inventory inputs:
_quarto.yml (registered notebooks + render config),
index.qmd, references.bib, _freeze/, data/, images/, tables/.
Phase 2 — Announce scope (do not wait — read-only)
State what will be audited (manuscript vs <slug>) and which dimensions apply.
Phase 3 — Audit dimensions
Run the dimensions from references/audit-checklist.md, recording each issue with
severity, location, and a concrete fix. Notebook scope runs only 1–4; manuscript
scope runs all:
- Freeze freshness — each registered notebook's
.qmd mtime vs its
_freeze/ cache: Current / Stale / Unexecuted.
- Data-path existence — grep notebook load calls, resolve relative to
notebooks/, classify Resolved / Broken / Undocumented against data/.
- Citation cross-check —
@key/[@key] in index.qmd (ignoring @sec-,
@fig-, @tbl-, emails) vs references.bib: missing / orphaned / duplicate.
- Figure/table export presence — every
{{< embed notebooks/...#fig-* >}}
and {{< include tables/*.md >}} target exists on disk.
- Placeholders —
[FILL: and [CITE: left in body prose.
- Anonymization — author names, self-citations, identifying acknowledgments,
repo URLs in the body (HIGH only if the venue is double-blind).
- Cross-references —
@sec-/@fig-/@tbl- targets resolve.
- Word count — prose-body word count vs typical 8,000 / 10,000 limits (INFO).
Phase 4 — Produce report
Use the format in references/scoring-and-criteria.md (verdict, dimension table,
issues table, manual submission checklist, priority action items). Deliver inline
and save it, creating the parent folder first (mkdir -p notes/<...>/):
- Notebook scope →
notes/<slug>/<slug>_review.md
- Manuscript scope →
notes/manuscript/manuscript_review_<YYYYMMDD>.md
Follow-up
Offer to: elaborate on any issue; run /project:execute to refresh stale
notebooks; run /project:literature check for the authoritative citation pass;
or apply fixes (with explicit go-ahead, since this skill is otherwise read-only).
