The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the vendor's CLI tool (pragman) to perform configuration and diagnostics. Instructions restrict execution to specific, controlled subcommands like scan, init, and doctor with JSON output for deterministic processing.
[EXTERNAL_DOWNLOADS]: The skill mentions the @prag-man/pragman-exp Node.js package. It includes an explicit safety instruction to offer the manual installation command to the user but forbids automatic execution by the agent.
[PROMPT_INJECTION]: The skill incorporates strong defensive measures against indirect prompt injection. It explicitly mandates treating all discovered skill content as untrusted data and strictly prohibits the agent from executing, importing, or following any instructions found within scanned files.

pragman-init