adversarial-review
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONNO_CODE
Full Analysis
- [DATA_EXFILTRATION]: The skill's primary instructions require the agent to send code and context to external models (GPT/Gemini). This process involves the movement of intellectual property or sensitive source code from the local environment to third-party service providers.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core function is to process and review external code. Malicious instructions could be embedded within the code changes or comments to manipulate the agent's review logic or behavior.
- Ingestion points: User-provided code changes identified for review.
- Boundary markers: None identified; the skill does not specify the use of delimiters to isolate untrusted code from the agent's primary instructions.
- Capability inventory: The skill contains no executable scripts, tool definitions, or automated network capabilities.
- Sanitization: No validation or sanitization of the input code is prescribed before it is processed or sent to external models.
- [NO_CODE]: The skill consists entirely of markdown instructions and configuration metadata. It does not include any executable scripts (.py, .js, .sh), binaries, or platform-specific tool definitions, which limits the risk of direct malicious execution.
Audit Metadata