| name | analyze-url |
| description | Deep analysis of a URL — tech stack, architecture, business model, competitive positioning. Triggers: "analyze url" | "deep dive url" | "analyze this site" | "analyze https://" | "what tech does this site use" | "reverse engineer this site" | "analyze this website" | "look into this site". |
Analyze URL
Scrape URL → deep technical and strategic analysis.
Entry
/analyze-url https://example.com
¬U → ask user to get one.
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 "$PLUGIN_ROOT" && uv run python scripts/doctor.py
- exit 1 → show output, stop, guide install. Optional warnings → inform user, continue.
- Skip on subsequent invocations.
Step 2 — Scrape
cd "$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"
Step 3 — Deep Analysis
SaaS / Product URLs
- Product overview — what, audience, value prop
- Tech stack signals — frameworks, infra, CDN, visible APIs
- Architecture patterns — SPA vs SSR, API-first, microservices signals
- Business model — pricing, freemium, enterprise, self-serve
- Competitive positioning — market category, differentiation, moat
- Growth signals — community size, hiring, funding, integrations
- Strengths (3-5 bullets) | Weaknesses (3-5 bullets)
GitHub Repos
- Project health — stars, forks, recent activity, contributor count
- Architecture — language, framework, project structure
- Code quality signals — CI/CD, tests, linting, type safety
- Documentation — README quality, docs site, examples
- Community — issues, PRs, discussions activity
- Notable patterns worth adopting
YouTube Videos
Run full video analysis pipeline (tempfile per ${CLAUDE_PLUGIN_ROOT}/../shared/references/tempfile-convention.md):
TMPDIR=$(mktemp -d -t "web-intel-video-analysis-XXXXXX")
trap 'rm -rf "$TMPDIR"' EXIT
ANALYSIS="$TMPDIR/analysis.json"
cd "$PLUGIN_ROOT" && SSL_CERT_FILE=/etc/ssl/certs/ca-certificates.crt uv run python scripts/video_analyzer.py "$URL" --output "$ANALYSIS"
Pipeline: scrape metadata + transcript → download video (yt-dlp, 1080p max) → extract frames at 1fps (ffmpeg) → auto-detect best local VLM by GPU VRAM (qwen3-vl via Ollama) → batch-describe frames → output JSON.
Read "$ANALYSIS" → analyze:
- Video overview — title, channel, duration, views, engagement rate
- Content analysis — thesis, structure, rhetorical techniques (from transcript)
- Visual analysis — techniques (3D/2D/photography), palette, composition, transitions (from frames)
- Production quality — animation, visual variety, text/typography, pacing
- Scene breakdown — visual timeline mapped to narration
- Strengths (3-5) | Weaknesses (3-5)
video_analyzer fails (¬GPU ∨ ¬Ollama) → fall back to transcript-only via standard scraper.
Articles / Content
- Thesis — main argument or claim
- Evidence quality — data, sources, methodology
- Implications — what this means for the field
- Counter-arguments — what the author missed
- Actionable insights
Step 4 — Present
Output structured markdown with clear headers. Include confidence (high|medium|low) for inferred info.
success: false → fall back to WebFetch.
$ARGUMENTS
