adversarial-review

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by instructing a sub-agent to process untrusted code retrieved from the local repository.
  • Ingestion points: The skill reads external data (code changes) via git diff and git status output.
  • Boundary markers: The dispatch instructions for the sub-agent do not specify the use of delimiters (e.g., XML tags, triple backticks) or explicit instructions to treat the ingested code as data rather than instructions.
  • Capability inventory: The skill executes shell commands using the git CLI and dispatches a secondary agent (adversarial-reviewer).
  • Sanitization: No content validation, escaping, or filtering is performed on the code content before it is passed to the sub-agent for analysis.
  • [COMMAND_EXECUTION]: The skill triggers standard version control commands (git status, git diff) to assess the repository state. These operations are expected for the skill's primary purpose of performing a code review and involve well-known developer tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 03:29 PM
Security Audit — agent-trust-hub — adversarial-review