The Agent Skills Directory

[SAFE]: The skill instructions and metadata do not contain any malicious patterns, command execution, or network requests.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it processes untrusted user data. Ingestion points: arguments and critique_set in SKILL.md. Boundary markers: Absent. Capability inventory: No tools, network operations, or file system access identified. Sanitization: Absent. The risk is minimized as there are no exploitable capabilities available to the skill.
[NO_CODE]: The skill consists exclusively of logical reasoning instructions and natural language processing schemas, with no external dependencies or executable scripts.

adjudicate-exchange