good-chad

Warn

Audited by Gen Agent Trust Hub on Oct 11, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains a bash code block intended for the agent to execute to check for an external configuration directory: [ -n "$POWERSUIT_OVERLAY" ] && ls -1 "$POWERSUIT_OVERLAY/stance" 2>/dev/null || true.
  • [DATA_EXFILTRATION]: The skill uses the shell command mentioned above to discover the contents of a directory specified by the $POWERSUIT_OVERLAY environment variable, which constitutes local file system discovery.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests user input through specific triggers (e.g., "good chad mode") to activate behavioral stance guidelines. These instructions prioritize specific model behaviors (e.g., "Hold position under pushback") that could influence the agent's responsiveness to user steering, creating an attack surface for behavioral manipulation. The skill lacks explicit boundary markers for these instructions when applied to user-provided data.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 11, 2026, 12:50 PM