benchmark

Benchmark

Let: U := target URL F := focus area (ui | features | stack | performance | ux | all), default all AB := agent-browser (preferred — adds snapshot -i for interactive refs) PW := uv run python scripts/screenshot.py (fallback — reuses web-intel's Playwright extra)

Scrape U → screenshot → analyze repo → compare → report.

Entry

/benchmark https://example.com             → full benchmark
/benchmark https://example.com --focus ui   → UI/design focus

¬U → → DP(B) Parse --focus <area> if provided.

Step 1 — Locate Plugin

PLUGIN_ROOT=$(find ~/projects -maxdepth 4 -path "*/web-intel/pyproject.toml" -print -quit 2>/dev/null | xargs dirname)
if [ -z "$PLUGIN_ROOT" ]; then
  echo "ERROR: web-intel plugin not found. Install: claude plugin install web-intel"
  exit 1
fi

First Use

First invocation in session only:

cd "$PLUGIN_ROOT" && uv run python scripts/doctor.py

• exit 1 → show output, stop, guide install. Optional warnings → inform user, continue.
• Skip on subsequent invocations.

Step 2 — Scrape Target

cd "$PLUGIN_ROOT" && SSL_CERT_FILE=/etc/ssl/certs/ca-certificates.crt REQUESTS_CA_BUNDLE=/etc/ssl/certs/ca-certificates.crt uv run python scripts/scraper.py "$URL"

Parse JSON: data.text (features/copy/value props), data.title, data.description, content_type. success: false → fall back to WebFetch.

Step 3 — Screenshot

Tempfile per ${CLAUDE_PLUGIN_ROOT}/../shared/references/tempfile-convention.md:

TMPDIR=$(mktemp -d -t "web-intel-benchmark-XXXXXX")
trap 'rm -rf "$TMPDIR"' EXIT
SHOT="$TMPDIR/screenshot.png"

∃ AB:

agent-browser open "$URL" && \
agent-browser wait --load networkidle && \
agent-browser screenshot --full "$SHOT" && \
agent-browser snapshot -i

¬∃ AB → PW fallback (web-intel already ships Playwright):

cd "$PLUGIN_ROOT" && uv run python scripts/screenshot.py "$URL" "$SHOT"

Neither available → skip UI/Design dimension, note in report (other dimensions still scored).

Step 4 — Analyze Repo

Use Glob + Grep to build capability inventory: pages/routes, API modules, UI components, auth flows, i18n, email templates.

Step 5 — Compare & Score

Build comparison matrix based on F:

Dimension	Weight	What to compare
Features	25%	Feature parity — what they have vs what we have
UI/Design	20%	Component richness, design system, visual polish
Stack	15%	Tech stack modernity, DX, build tooling
Auth & Security	15%	Auth methods, RBAC, security headers
Performance	10%	SSR, caching, bundle optimization signals
UX Patterns	15%	Loading states, error handling, a11y, i18n

∀ dimension → assign: Ahead | Parity | Gap | Missing

Step 6 — Present Results

Output structured markdown: Target Overview → Comparison Matrix → Overall Score (weighted %) → Gap Analysis → Priority-ordered Recommendations.

→ DP(A) Create issues for gaps | Deep-dive a dimension | Benchmark another URL | Done

Safety Rules

• Read-only repo analysis (no modifications)
• Single scrape per URL per invocation
• No credentials/tokens stored from scraped sites

$ARGUMENTS