Review

Universal Review

Auto-detect file format and launch the right review agents in parallel. Read-only — no fixes applied.

Format Detection

Determine file type from extension:

• .tex → Beamer LaTeX slides
• .qmd → Quarto RevealJS slides
• .typ → Typst document (slides, docs, guides, CVs)
• .py → Python script
• .md → Markdown

For .typ files, also detect document type from content:

• Contains #import "@preview/polylux" or #import "@preview/touying" → Typst slides
• Contains #import "@local/qk" → Uses qk component library
• Contains #set page(width: 16cm, height: 9cm) or similar → Presentation
• Otherwise → Typst document (guide, essay, CV, etc.)

Routing Table

| Extension | Agents (parallel) | Condition | |-----------|-------------------|-----------| | .tex | proofreader + document-auditor + pedagogy-reviewer | Always | | .tex | + diagram-reviewer | If TikZ found (\begin{tikzpicture}) | | .qmd | proofreader + document-auditor | Always | | .qmd | + quality-critic | If .tex sibling exists | | .typ | typst-reviewer + proofreader | Always | | .typ | + pedagogy-reviewer | If slide format detected | | .py | python-pro (subagent) | Always | | .md | proofreader | Always |

Steps

1. Parse $ARGUMENTS — resolve file path
2. Detect format from extension
3. For .typ files, read file to detect document subtype
4. Launch agents in parallel per routing table
5. Collect all reports from quality_reports/
6. Synthesize a consolidated summary table

Output

# Review Summary: [Filename]

| Dimension | Agent | Critical | Major | Minor | Verdict |
|-----------|-------|----------|-------|-------|---------|
| [varies] | [agent name] | N | N | N | [status] |

## Overall: [EXCELLENT / GOOD / NEEDS WORK / POOR]

## Critical Issues (if any)
[List top issues requiring immediate attention]

## Reports Generated
- quality_reports/[file]_report.md
- quality_reports/[file]_typst_review.md
- ...

Quality Score Rubric

See typst/references/quality-gates.md for the authoritative score rubric, severity definitions, and dispatch table.

Important

• This is a read-only review — no fixes applied
• For review + fix, use /finish instead
• For format-specific deep review, use the individual agents directly

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: clearsmog
• Source: clearsmog/claude-skills
• License: MIT
• Homepage: https://github.com/clearsmog/claude-skills

Install and usage instructions live in the source repository linked above.