adversarial-review

Fail

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use a specific mode to suppress security prompts.
  • Evidence: In the 'Permission Mode' section, the instructions state to 'always pass mode: "bypassPermissions"' to avoid 'permission prompts during unattended execution.'
  • This directive bypasses standard user-consent workflows and privilege controls within the agent execution environment.
  • [PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection through its primary function.
  • Ingestion points: The skill accepts a <file-or-directory> argument and reads the content of these files for review.
  • Boundary markers: Absent. There are no instructions or delimiters provided to ensure the agent ignores malicious commands that might be embedded in the code files being analyzed.
  • Capability inventory: The skill utilizes the Agent tool to spawn sub-agents and integrates with multiple review and validation tools.
  • Sanitization: Absent. The skill performs no validation or escaping of the external file content before it is processed by the model.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 15, 2026, 06:17 PM
Security Audit — agent-trust-hub — adversarial-review