/cs-md-review

disable-model-invocation: true
description: Convert a markdown PR writeup or code review (with ```diff blocks and > [!BLOCKER]/[!MAJOR]/[!MINOR]/[!NIT] severity callouts) into a single-file 2-column HTML review. Top jump-nav lists every finding by severity. Diff on the left, severity-tagged annotation cards on the right, mandatory named reviewer footer. Refuses without --reviewer (a code review must name a human) or if no diff hunks present.
argument-hint: "<path to markdown> --reviewer \"<name>\" [--title \"<PR title>\"] [--severity-convention \"<a,b,c,d>\"]"

/cs-md-review — Code-review markdown → 2-column HTML

Convert the review markdown at **$ARGUMENTS** into a single-file 2-column HTML review.

Pre-flight gates (refuse, never override)

1. **Input < 100 lines** → refuse (markdown wins below Shihipar's threshold). 2. **Design-system not onboarded** → refuse, surface `/cs-design-system`. 3. **`--reviewer` missing** → refuse (a code review must name a human reviewer). 4. **No diff hunks present** → refuse, route to `md-document` instead. 5. **Output directory unwritable** → refuse, ask user for alternate via `--out`.

Pipeline

# 1. Resolve output path (uses doctype=review → review- prefix)
python3 ${CLAUDE_PLUGIN_ROOT}/skills/markdown-html-orchestrator/scripts/output_path_resolver.py \
    --input "<path>.md" --doctype review

# 2. Parse diff hunks
python3 ${CLAUDE_PLUGIN_ROOT}/skills/md-review/scripts/diff_parser.py \
    --input "<path>.md" --output /tmp/hunks.json

# 3. Extract severity-tagged annotations, attached to nearest hunk
python3 ${CLAUDE_PLUGIN_ROOT}/skills/md-review/scripts/annotation_extractor.py \
    --input "<path>.md" --diff-blocks /tmp/hunks.json --output /tmp/annotations.json

# 4. Render — --reviewer is mandatory
python3 ${CLAUDE_PLUGIN_ROOT}/skills/md-review/scripts/review_html_renderer.py \
    --diff-blocks /tmp/hunks.json --annotations /tmp/annotations.json \
    --reviewer "<reviewer-name>" --title "<PR title>" \
    --output <resolved-out>.html

What ships in the HTML

• **Top jump-nav** — every finding with severity badge (color + icon + aria-label) + 80-char preview + jump link; counts in the heading ("3 BLOCKER · 2 MAJOR · 1 NIT")
• **2-col hunk rows** — unified diff on left (line numbers both sides, +/− marks, addition/deletion bg tints from design-system tokens), annotation cards on right
• **Severity coding** — BLOCKER = computed danger color (accent rotated 120° toward red), MAJOR = `--md-warn`, MINOR = `--md-link`, NIT = `--md-text-muted`. Every badge has icon (■ / ▲ / ● / ◦) + aria-label per WCAG 1.4.1
• **Approval bar** — if LGTM markers and no findings, success-tinted bar
• **General-comments section** — for unanchored annotations
• **Reviewer footer** — "Reviewer: \<name\>" (mandatory)
• **Responsive** — 2-col collapses to stacked under 900px

Hard rules

• Output is one `.html` file. No external CSS/JS. Only external is Google Fonts CSS (no Prism — diff colors conflict with syntax highlighting).
• No JS framework runtime. The page is fully static; jump-nav links are plain anchors.
• Severity is never color-only (WCAG 1.4.1).
• Re-running on the same input writes `review-{slug}-2.html` etc.

Custom severity convention

--severity-convention "critical,important,suggestion,nit"

Swaps tier names; position 0 is most severe. Default: `BLOCKER,MAJOR,MINOR,NIT` (Google *Code Review Developer Guide*).

Output

Returns: hunk count, annotation count, severity breakdown, output path, reviewer name, one forcing question.

References

See `markdown-html/skills/md-review/references/`:

• `diff_rendering_canon.md` — POSIX diff format + GitHub/GitLab conventions + difftastic + SWE at Google
• `severity_coding.md` — WCAG 1.4.1 + Google review taxonomy + Don Norman signifier
• `pr_annotation_ux.md` — Convergent 2-col UX from GitHub/GitLab/Reviewable/CodeStream