The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script located at team/cortex/scripts/cortex_agent/eval_scan.py to perform static analysis on the codebase. This is a primary function of the skill and intended for the user's workspace.
[DATA_EXPOSURE]: The skill scans for ML-related artifacts and reads configuration files like requirements.txt and pyproject.toml. This behavior is limited to the local environment and intended for environment detection to help the ML engineer agent understand the stack.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from untrusted content in project files (logs, configuration) during its evaluation steps.
Ingestion points: Project configuration files (requirements.txt, pyproject.toml) and log files (metrics/, logs/) are read into the agent context in Steps 1, 2, and 3.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when reading these files.
Capability inventory: The agent has access to Bash, Write, and Edit tools which provide a high capability tier for potential exploitation.
Sanitization: No explicit sanitization or filtering of file contents is performed before processing.

cortex-eval