The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill directs users to download the official Claude Code CLI from the trusted domain claude.ai and provides links to official documentation hosted by Anthropic.
[REMOTE_CODE_EXECUTION]: Provides the official installation command (curl | bash) for the Claude Code CLI. Since this targets a well-known service domain, it is considered safe for the intended development context.
[COMMAND_EXECUTION]: Examples demonstrate how to build agents that execute system commands while explicitly providing templates for implementing security hooks to validate and restrict these commands (e.g., blocking rm -rf or access to .env files).
[SAFE]: The skill emphasizes the use of built-in authentication via ~/.claude/ instead of hardcoding API keys and recommends budget limits and turn constraints to prevent runaway execution costs.

agent-builder