The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional Markdown documentation and JSON metadata designed to guide developers in building subagents for Claude Code.
[COMMAND_EXECUTION]: The documentation provides examples of using local shell scripts for lifecycle hooks (e.g., ./scripts/validate-readonly-query.sh) and standard Git commands for worktree isolation. These are presented as developer-managed workflow and security features.
[EXTERNAL_DOWNLOADS]: The .dev/reference-inventory.json file lists third-party technical blogs (shipyard.build, smartscope.blog) and a community GitHub repository (VoltAgent) for supplementary reading. These are informational links and do not involve automated code execution.
[DYNAMIC_EXECUTION]: The documentation explains the Claude Code "Hooks" and "MCP Servers" systems, which allow the platform to execute local commands or connect to external services. A sample defensive script is provided to demonstrate how to validate and block destructive database operations.

subagent-engineering