cross-eval
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill communicates with well-known external AI services (OpenAI/Codex and Gemini) to provide comparative analysis. This involves sending user-provided document content to these providers when the corresponding API keys or CLI tools are configured in the environment.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text in the form of business memos or briefs. Since this content is prepended with a prompt and sent to multiple LLMs, there is a theoretical surface for indirect prompt injection if the source document contains instructions meant to manipulate the reviewer persona. However, this is inherent to any tool processing external natural language.
Audit Metadata