SteveLeve

cloudflare-docs-specialist

Use Cloudflare MCP servers to search and quote official Cloudflare docs; prefer authoritative pages for Workers, AI, D1, R2, KV, Vectorize, Queues, and deployment guidance.

SteveLeve 1 Updated 1mo ago
GitHub

Install

npx skillscat add steveleve/chatbot-demo-cloudflare/cloudflare-docs-specialist

Install via the SkillsCat registry.

SKILL.md

Cloudflare Docs Specialist

Use this skill when a question needs Cloudflare documentation, product specifics, or the latest platform behavior.

When to Trigger

  • User asks “how do I… on Cloudflare/Workers/AI/Gateway/Vectorize/D1/R2/KV/Queues/DO/Pages”.
  • Need authoritative limits, flags, config keys, or migration steps.
  • Need to cite docs or compare old vs new behavior.

Tools

  • mcp__CloudflareBindings__search_cloudflare_documentation (primary; returns doc chunks). If unavailable, fall back to local repo docs under docs/.
  • Prefer multiple queries if first result looks generic.

Workflow (tight loop)

  1. Clarify product & operation (e.g., “Workers AI embeddings via AI Gateway”).
  2. Run targeted search with product keyword + task (e.g., "AI Gateway cache TTL" or "D1 batch"), limit to few results.
  3. Read snippets; open more if needed; avoid loading entire manuals.
  4. Synthesize concise answer with cites; include version/flag names and defaults; note remote-only constraints for Workers AI/Vectorize.
  5. Record key command/config snippets that fit the repo (Hono + wrangler). Keep it short.

Repo-aware defaults

  • Use wrangler dev --remote for AI/Vectorize/D1 in this project.
  • Bindings live in wrangler.jsonc; cite binding names as seen there.
  • Link GitHub issues when relevant (#12–#19, #26–#27).

Style

  • Lead with the concrete steps; then add cautions/limits.
  • Cite sources; prefer official docs over blogs.
  • Keep answers under 200 tokens unless asked otherwise.