Skip to main content

Bundled Skills

Bundled skills ship inside Claude Code, so you need no files to use them. They are prompts that guide Claude through a common job. For example, /simplify reviews changed code and fixes the problems it finds.

You cannot change a bundled skill, but you can replace one by making your own skill with the same name. Turn them off with the disableBundledSkills setting or CLAUDE_CODE_DISABLE_BUNDLED_SKILLS=1.

Example​

claude -p "/debug the login test fails on CI"

/debug turns on debug logging and then looks into the issue you describe.

Controls​

  • The setting disableBundledSkills and the env var CLAUDE_CODE_DISABLE_BUNDLED_SKILLS remove them. /doctor stays available.
  • When the policy allow_bundled_skills is not granted, only names in a fixed list stay. The list in the program is: artifact-diagramming, batch, claude-code-docs, code-review, commit, dataviz, debug, explain-usage, fewer-permission-prompts, keybindings-help, loop, memory-types, plugin-authoring, pr, simplify, update-config, verify, workflow-authoring.

Skills​

SkillStatusWhat it doesNotes
/doctor (alias /checkup)DocumentedChecks your setup: install, unused extensions, CLAUDE.md, hooks, permissions. /doctor prompt-audit audits instructions.Hidden by env var DISABLE_DOCTOR_COMMAND or skillOverrides "doctor": "off". Needs a workspace. Interactive.
/claude-apiDocumentedA reference for the Claude API and Managed Agents in your language.Subcommands: migrate, upgrade, managed-agents-onboard, prompt-audit, cost-optimize, build-eval, hillclimb, preserved-thinking-migration, managed-agents-secure.
/claude-api managed-agents-secureDocumentedCompares a Managed Agents agent's settings with Anthropic's published security guide. It changes nothing.New in 2.1.304. It is unavailable when Claude Code's check on its ant commands does not run. Source: the release notes and the program.
/debug [issue description]DocumentedTurns on debug logging and investigates.disableModelInvocation. Allowed tools: Read, Grep, Glob.
/batch <instruction>DocumentedPlans a large change and runs 5 to 30 worktree agents that open pull requests.disableModelInvocation.
/runDocumentedLaunches your app so you can see a change work.Reads a recorded run-<name> skill if one exists.
/run-skill-generatorDocumentedRecords how to run the project as .claude/skills/run-<name>/.disableModelInvocation.
/verifyDocumentedBuilds and runs the app to check a change.Records its recipe in .claude/skills/verify/SKILL.md.
/code-review (alias /review), /loop, /simplify, /workflow-authoringDocumentedNamed in the docs./workflow-authoring needs dynamic workflows.
/plugin-authoringDocumentedHelps you write and share a mod or plugin.Since 2.1.304, sharing no longer stops at the secret scan on machines with managed settings or on Team and Enterprise plans. Source: the release notes.
/update-configDocumentedEdits settings.json: hooks, permissions, environment variables.Registered as a bundled skill.
/fewer-permission-promptsDocumentedScans transcripts and adds a read-only allowlist to .claude/settings.json.Needs a workspace.
/security-reviewDocumentedReviews pending branch changes for security risks.A prompt command that Claude can also call through the Skill tool. Git repositories.
/schedule (alias /routines)DocumentedCreates and manages scheduled remote agents.A stub with a name, alias, and description. Needs claude.ai.
/keybindings-helpUndocumentedHelps you edit ~/.claude/keybindings.json.userInvocable:false, so only Claude runs it. Allowed tool: Read.
/claude-in-chromeGatedBrowser automation through the Chrome extension.No allowed tools.
/design-sync [<project hint>]GatedPushes design-system components to claude.ai/design.disableModelInvocation.
/artifact-componentsGatedReusable components in an Artifact.

Source: Commands, Extend Claude with Skills, and the program (the list of bundled skill names and the allow_bundled_skills key).