The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the use of shell commands for development automation. It includes specific guidance on limiting tool access to only necessary commands (e.g., using allowed-tools: Bash(git:*)), which is a recommended security practice to reduce the attack surface.
[DYNAMIC_CONTEXT_INJECTION]: The documentation explains the !command syntax for injecting dynamic system context into command prompts. The examples provided (such as git status, gh pr view, and npm test) are benign and appropriate for the intended development context.
[INDIRECT_PROMPT_INJECTION]: The skill teaches how to create commands that accept user-supplied arguments. It provides a 'Validation Patterns' section that demonstrates how to sanitize these inputs using shell utilities like grep and test, effectively providing developers with the tools to mitigate injection risks.
[DYNAMIC_EXECUTION]: The skill describes the use of ${CLAUDE_PLUGIN_ROOT} for referencing plugin-internal scripts and resources portably. This is a standard architectural pattern for plugin development and does not facilitate the execution of untrusted external code.

command-development