菜单
Orchestrator (Grok 4.5, Fable 5, Sonnet 5, or GPT-5.6 Sol) with Composer 2.5 workers. Two separate setups: Cursor IDE (subagents in .cursor/agents) or Claude Code (cursor-agent CLI + .claude/skills). Use when configuring orchestration; pick the guide for your host.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | agent-orchestration |
| description | Orchestrator (default Grok 4.5) with composer-2.5[fast=false] workers. Two separate setups: Cursor IDE (subagents in .cursor/agents) or Claude Code (cursor-agent CLI + .claude/skills). Use when configuring orchestration; pick the guide for your host. |
| user-invocable | true |
| disable-model-invocation | true |
Goal: Grok 4.5 (default orchestrator) plans; composer-2.5[fast=false] executes bulk work cheaper.
Install once, then run only the init for your host from .agents/skills/agent-orchestration/:
bunx skills add FixMyBerlin/fixmyskills --skill agent-orchestration -a cursor -y
| Host | Mechanism | Files | Bootstrap | Guide |
|---|---|---|---|---|
| Cursor IDE | subagents (model: pins) + @orchestrator-worker | .cursor/agents/, .cursor/rules/ | init-cursor.sh | cursor-ide.md |
| Claude Code | cursor-agent CLI (cursor-agent on PATH) | .claude/skills/cursor-worker-* + CLAUDE.md snippet | init-claude.sh | claude-code.md |
Do not mix both inits unless you use both tools on the same repo. Cursor personal workers: cursor-personal-setup.md.
Orchestrate greenfield FixMyBerlin/FMC repo bootstrap: plan which fixmyskills to install, git init, AGPL license, three finish-work commits. Use when scaffolding a new project or onboarding a new repo. Delegates to finish-work, tech-stack, and agent-orchestration — do not duplicate their steps.
nuqs (type-safe URL query state) for Next.js and legacy/shared components. Prefer TanStack Router validateSearch on TanStack Router/Start apps. Use when writing nuqs hooks, parsers, NuqsAdapter setup, or Next.js URL state — not for greenfield TanStack route search params.
E2E testing and browser automation for TanStack Start apps (FMC/TILDA). Use @playwright/test for committed suites/CI; agent-browser MCP (skill tech-stack) for agent exploration; this skill for smoke tests, stubbed auth, map hooks, and quick /tmp scripts. Not for Next.js.
React 19 + TypeScript for FMC TanStack Start apps: typed components, events, generics, ref-as-prop, Compiler-first memoization, and useEffect discipline (naming, when not to use Effect, alternatives). Use when typing React components/hooks/events, or writing/reviewing useEffect, useEffectEvent, derived state, data-fetch effects, or eslint-plugin-react-hooks effect rules. Not for routing/server boundaries (see tanstack-start-* skills).
react-map-gl/maplibre patterns for FMC geo apps: declarative Source/Layer (never addSource/addLayer), MapProvider + useMap(), interactiveLayerIds, flat Source/Layer, map-loaded guard, window.__mainMap debug exposure, Map event handlers (not useEffect), URL map state, initialViewState, cursor, feature state, missing images. Use when building or reviewing Map components, layers, click/hover handlers, setFeatureState, map URL sync, or agent/test map inspection in TanStack Start / Vite apps. Reference implementation: tilda-geo/app.
Review and merge open Dependabot PRs in the current workspace repo: changelog triage, risk classification, CI verification, and rebase merge. Use for "/review dependabot" or similar — always scoped to the repo you have open. Weekly Bun/GitHub Actions/Docker bumps, grouped patches, security updates, semver-major upgrades.