skills/wcof/prdcontextengine/prd-scan/Gen Agent Trust Hub

prd-scan

Fail

Audited by Gen Agent Trust Hub on May 10, 2026

Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill description explicitly states its purpose is to "scan all AI tool project sessions and batch collect" them. This indicates a high risk of unauthorized harvesting of sensitive conversation history, internal project context, and proprietary data from the user's environment.
  • [COMMAND_EXECUTION]: The skill bundle includes several files (checks, commands, modules, scripts, support) that contain path traversal strings ("../../"). This pattern is used to escape the skill's local directory and access the file system at higher levels, potentially reaching internal platform scripts or other users' data.
  • [COMMAND_EXECUTION]: The skill instructions direct the agent to dynamically search for and execute a Python script (prd-command-dispatch.py) if the primary path fails. This behavior encourages the agent to bypass restricted paths to find and run unverified code located elsewhere on the system.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 10, 2026, 04:23 AM