The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local Python scripts to perform diagnostic tasks required for evaluation.
scripts/check_skill_quality.py executes py_compile via subprocess.run to verify that Python scripts within the evaluated skill directory are syntactically valid.
scripts/discover_agent_runtime.py executes common agent CLI commands (such as codex --help and claude --version) to determine available tracing and evidence capture capabilities in the execution environment. These operations are limited to informational help flags and do not involve shell injection or user-supplied command strings.
[SAFE]: The skill's operations are well-scoped and focus on local file analysis and system discovery.
No hardcoded credentials, sensitive file access, or network exfiltration patterns were detected.
The skill includes explicit 'Operating Rules' and an 'Iron Law' designed to ensure the agent evaluates skills based on raw evidence rather than instructions alone, which serves as a safeguard against indirect prompt injection.
All dependencies are from the Python standard library.

skill-evaluator