The Agent Skills Directory

[SAFE]: The skill's operations, including codebase analysis and guideline generation, are consistent with its documented purpose. No malicious patterns such as remote code execution, exfiltration, or persistence were found.
[PROMPT_INJECTION]: The skill processes untrusted codebase data and feedback, creating an indirect prompt injection surface. This risk is addressed through the mandatory use of the TC-EBC (Task, Context, Elements, Behavior, Constraints) framework, which enforces explicit boundaries and constraints to ensure the agent ignores instructions embedded in external data.

loom