| name | site-extract |
| description | Extract the smallest useful amount of text or structured data from a known public page before falling back to full browser automation. |
| origin | repo-local web research workflow |
Site Extract
Use this skill when the target page or docs site is already known and the task is to extract facts, structured fields, or bounded excerpts.
Workflow
- Prefer an official API, feed, raw markdown, or direct HTML fetch over a rendered page.
- Extract only the sections needed for the task instead of dumping whole pages.
- Keep the source URL, page title, publish/update date, and access date with the notes.
- Normalize findings into concise bullets, tables, or JSON-shaped notes.
- If the content depends on client-side rendering or interaction, switch to
browser-fallback.
Rules
- Prefer text or DOM extraction over screenshot-only reading.
- Preserve enough structure for later citation and verification.
- Do not copy long copyrighted passages when a short summary is enough.
- If the site exposes stable machine-readable fields, use those rather than re-parsing visible prose.
Output
- Bounded structured notes tied to exact URLs.
- Missing fields or extraction blockers called out explicitly.
Read next
docs/reference/ai-agent-web-research-stack.mdbrowser-fallback or source-audit when needed
