The Agent Skills Directory

[SAFE]: The skill provides architectural and security guidance for a specific authentication protocol. The instructions mandate defensive behaviors, such as requiring a safety verdict before code generation and forbidding the agent from collecting or handling user passwords.
[EXTERNAL_DOWNLOADS]: The scripts/probe_auth_md.py script performs HTTP GET and HEAD requests for protocol discovery. It implements multiple safety controls, including URL scheme validation, blocking private and special-use IP ranges by default, and enforcing strict response size limits to mitigate SSRF and resource exhaustion attacks.
[COMMAND_EXECUTION]: The evals/test_probe.py script uses subprocess.run to execute the local discovery script for unit testing. The implementation is secure as it avoids shell execution and uses a fixed list of arguments, ensuring the command execution is scoped to the skill's own testing procedures.

auth-md