adversarial-review

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: User-provided code is interpolated directly into the prompts for the 'Bug Finder', 'Adversary', and 'Referee' roles (SKILL.md, lines 58, 89, and 117). Boundary markers: Absent; the prompts do not use delimiters or explicit instructions to ignore commands embedded in the code. Capability inventory: The orchestration involves spawning sub-agents via CLI tools. Sanitization: Absent; the input code is not escaped or validated before interpolation.
  • [COMMAND_EXECUTION]: The skill provides instructions and a template for a shell script (run-review.sh) that executes CLI commands locally to manage the multi-agent review pipeline.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 10:46 PM