anthropic-platform-expertlisted

# Anthropic Platform Expert Skill Use this skill when Alan asks about the Agents SDK, Claude API / Messages API, `docs.anthropic.com`, Claude Code headless (`claude -p`), SDK sessions/subagents/stream events, structured output, telemetry, official changelogs, or public status updates (@Anthropic / @claudeai) as they affect bridge/runner parity or OAuth-only evidence. **Do not use this skill** when Alan says `runner-command-builder`, `commands only`, or wants paste-ready `node bin/local-bridge-runner.js` lines — that is **runner-command-builder** (Terminal commands only; no lab-notes edits in Cursor). For policy, Terms, subscription vs API, or June 15 billing posture, use **anthropic-official** instead. ## When NOT to use (mandatory) - Alan said **runner-command-builder**, **commands only**, or wants **paste-ready Terminal CLI** — use `.cursor/skills/runner-command-builder/` instead (that skill does **not** edit files; it only prints `node bin/local-bridge-runner.js` blocks). - The message is only about how to invoke the runner, not researching Anthropic docs. ## How lookups work (plain language) When Alan asks a technical Anthropic question, **fetch the official page first** (WebFetch on a Tier-1 URL). Use WebSearch only to **find** the right official URL or a changelog/status post. **Context7 (ctx7) is not the default** — use it only if the fetch failed, the page was incomplete, or Alan explicitly wants indexed API snippets. ## Rules Playground only: `/Users/alanman