菜单
Use when unsure which Claude Code capability fits a task, when about to single-thread large/cross-cutting/repetitive work, or when you or the user ask what Claude Code can do — loads the full capability map (signal → capability) and points to the deep-dive skills.
基于 SOC 职业分类
Use when building custom AI agents headlessly, embedding Claude Code tools in an app, or running scripted/cron agents—need SDK package names, setup, core API entry points, supported languages, and relation to Claude Code CLI and Messages...
Use when needing to understand or manage background agents, long-running tasks, polling, scheduling, or async work in Claude Code without blocking the session.
Use when you need to undo code edits, explore alternatives without losing a starting point, recover from a bad edit path, or restore previous conversation state — press Esc Esc or run /rewind to open checkpoints menu.
Use when a task is too big for one context — broad audits, multi-file migrations, multi-source research, or parallel multi-dimension review — and you want to fan out and orchestrate many subagents deterministically with the Workflow tool (requires ultracode / explicit opt-in).
Use when deciding how to work on a task - choosing effort level, model, or fast mode to balance reasoning depth, speed, and token cost.
Use when setting up or debugging Claude Code GitHub Actions automation, or deciding whether to route CI/CD tasks to GitHub Actions vs local session.
| name | claude-code-capabilities |
| description | Use when unsure which Claude Code capability fits a task, when about to single-thread large/cross-cutting/repetitive work, or when you or the user ask what Claude Code can do — loads the full capability map (signal → capability) and points to the deep-dive skills. |
| user-invocable | true |
| argument-hint | (optional) a task or capability you want mapped, e.g. 'migrate 40 files' or 'hooks' |
This skill is your capability index. Its job: before you plan, single-thread something big, or decide a thing is impossible, make sure you have considered the right built-in Claude Code capability.
references/capability-map.md (in this skill's directory). It maps concrete task signals → the capability to reach for → how to invoke it → which deep-dive skill to read. It also lists the anti-patterns that this primer exists to prevent.references/capability-map.md)| Signal | Capability | Deep-dive skill |
|---|---|---|
| Repeatable workflow / knowledge you keep re-deriving | Skills & slash commands | claude-code-skills |
| Research, exploration, big review, 2+ independent tasks | Subagents (Agent/Task) | claude-code-subagents |
| Broad audit, migration, fan-out > one context | Dynamic workflows (Workflow, ultracode) | claude-code-dynamic-workflows |
| "Each time X, do Y" the harness must enforce | Hooks | claude-code-hooks |
| Long build/test/watch, polling external state | Background agents & tasks | claude-code-background-agents |
| Undo a bad edit path / restore state | Checkpointing & /rewind | claude-code-checkpointing |
| Live external tool / API / data source | MCP servers | claude-code-mcp |
| Reuse skills/hooks/agents/MCP as a bundle | Plugins | claude-code-plugins |
| Server-side PR / issue automation | GitHub Actions (@claude) | claude-code-github-actions |
| Build a custom/headless agent | Agent SDK | claude-code-agent-sdk |
| Task too hard/trivial for current setting | Effort, models & fast mode | claude-code-effort-models |
Each row's deep-dive skill is a separate skill in this plugin — invoke it with the Skill tool when you are actually doing that kind of work (they auto-surface on the matching signal too).
This primer keeps itself current:
hooks/auto-update.sh) does a throttled git pull --ff-only so your install tracks the latest merged main, and flags this card when your installed Claude Code is newer than the version the content was generated for (references/.generated-for).claude-code-* skills for any new capabilities, verifies every claim against the live docs, and opens a PR — auto-merging only when verification passes (AUTO_MERGE_POLICY = verified; see references/refresh-runbook.md).To refresh manually: run bash scripts/update-docs.sh, then follow references/refresh-runbook.md in a Claude session.