The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted application code, Docker Compose files, and CI pipelines as its primary input. This creates an attack surface where malicious instructions hidden in the analyzed data (such as code comments or configuration strings) could attempt to influence the agent's reporting behavior. The skill mitigates this by instructing the agent to only report findings with a confidence score of 7/10 or higher and by requiring manual verification of all evidence before the report is written to the file system.
[COMMAND_EXECUTION]: The skill executes a local Python script (fitness-config.py) located in the project's scripts directory to resolve configuration weights. The execution includes the target path as a command-line argument to allow for directory-specific overrides.

review-reliability