web-api-docs

Two-stage execution

This is a knowledge-base skill: 12k+ MDN docs are accessible via on-demand fetch, but only the ones relevant to the current conversation belong in your context.

How this invocation goes depends on whether arguments were passed:

• With arguments — explicit /web-api-docs <verb> [query]. The block below is the result of that command. Use it directly; no further extraction step needed.
• Without arguments — implicit auto-activation, or a bare /web-api-docs. The block below is guidance: instructions for you to scan the recent conversation, pick a query, and run find yourself. Follow it.

!`python3 ${CLAUDE_SKILL_DIR}/scripts/mdn.py $ARGUMENTS`

Subcommand cheatsheet

If the first argument isn't a known verb, all args route to find (so mdn.py css grid ≡ mdn.py find css grid).

Examples (after the initial guidance/result above)

Re-invoke this skill via its slash command — do not call the underlying script directly. On this install the command is /web-api-docs (it matches the directory holding this SKILL.md, so a renamed install would use that name instead).

/web-api-docs find intersection observer threshold
/web-api-docs find Array.flatMap
/web-api-docs get Web/CSS/:hover
/web-api-docs get https://developer.mozilla.org/en-US/docs/Web/API/fetch
/web-api-docs browse Web/API/Fetch_API

Issue multiple find invocations in parallel when the conversation touches multiple distinct topics — each is independent and cached separately.

Important notes

• Browser-compat tables, live samples, and interactive-example macros are replaced with > _(... omitted — see MDN.)_ placeholders. Standalone fenced code blocks (often labeled css interactive-example ```) remain in the output as useful example code. Don't fabricate compatibility data — link the MDN URL.
• The src: line under each get header tells you whether the body came from cache, a 304 revalidation, or a network fetch — surface that when freshness matters.
• For "what's the canonical URL of this doc" use the MDN: line in the header, or --json and read .url.
• 404 from get means the slug is wrong or moved without a redirect — fall back to find.
• find can return out-of-scope slugs (e.g. Learn/...) that aren't in our shipped index. get still works on those.
• For slug-encoding rules, macro substitution table, cache layout, and environment variables, see ${CLAUDE_SKILL_DIR}/reference.md.