| name | review-r |
| description | Review R scripts for code quality, reproducibility, and domain correctness. Checks set.seed, paths, structure, figure quality, and conventions. Use when the user asks to review R code or check scripts in scripts/R or Figures. |
Review R Scripts
Run the R code review protocol. Produce reports only — do not edit R files.
Steps
-
Identify scripts from the user's message (one file, "LectureN", or "all"). Look in scripts/R/, Figures/*/, and project .R files.
-
For each script, check (use .claude/rules/r-code-conventions.md if present):
- Structure & header: title, author, purpose, sections
- Reproducibility: single
set.seed() at top, relative paths, no absolute paths, dir.create() for outputs
- Console hygiene: no
cat()/print() for status; use message() sparingly
- Functions: snake_case, documentation, no magic numbers
- Domain correctness: estimator matches theory, correct standard errors, correct estimand (e.g. ATT vs ATE)
- Figure quality: transparent bg for Beamer, explicit dimensions, readable fonts
-
Save report to quality_reports/[script_name]_r_review.md. Use severity: Critical / High / Medium / Low.
-
Do NOT edit R source files. Present summary: issues per script, severity breakdown, top critical issues.
