The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bundled bash script scripts/repo-inventory to collect repository metadata. This script is read-only and uses standard diagnostic tools like git, tree, ls, and grep to identify languages, package managers, and task runners.
[DATA_EXPOSURE]: The inventory script specifically checks for the existence of .env.example and config/ directories to inform the user about setup requirements. It does not read the contents of sensitive files or attempt to exfiltrate data.
[PROMPT_INJECTION]: The skill uses instructional language such as 'CRITICAL' and 'IMPORTANT', but these are used as framing for the AI agent to prioritize actual code evidence over potentially stale documentation, which is a development best practice rather than an attempt to bypass safety filters.

agentsmd-generator