The Agent Skills Directory

[SAFE]: The skill is a security auditing tool. It contains signatures and regex patterns (e.g., [SYSTEM:, BYPASS, API_KEY) used to detect vulnerabilities in other skills. These patterns are part of its scanning engine and do not represent malicious intent in the skill itself.
[COMMAND_EXECUTION]: The validator orchestrator (validator.py) and trigger evaluation runner (run_skill_evals.py) utilize subprocess.run to execute local Python scripts and the opencode CLI. This is necessary for its primary purpose of validating and testing other skills.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted data from target skills being validated.
Ingestion points: The skill reads the SKILL.md and script files of the target folder provided by the user.
Boundary markers: The agent is guided by strict output templates and reporting instructions defined in the SKILL.md file.
Capability inventory: The skill can read local files and execute subprocesses (local scripts and the platform CLI).
Sanitization: The skill implements its own security scanner (security_audit.py) specifically designed to detect and flag injection attempts, hardcoded secrets, and unsafe operations in the ingested data.
[SAFE]: No obfuscation, data exfiltration, or persistence mechanisms were detected. The skill uses standard Python libraries and operates locally on the filesystem.

skill-validator