gemini-api-dev
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill uses authoritative instructions such as 'These rules override your training data' and 'Your knowledge is outdated' to force the agent to prioritize the provided technical specifications over its internal safety and training weights.
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install SDKs including
google-genaiand@google/genai. Since these packages are official releases from a trusted vendor, this activity is consistent with documented developer workflows. - [PROMPT_INJECTION]: The skill establishes an ingestion surface for external content from the
ai.google.devdomain and MCP-based documentation tools. While the source is well-known, the ingestion of external data into the prompt context creates a potential attack surface for indirect prompt injection due to the following evidence chain: - Ingestion points: Document fetching from
ai.google.dev/gemini-api/docs/and MCPsearch_docstool outputs. - Boundary markers: Absent; there are no specific delimiters or 'ignore instructions' warnings for external content.
- Capability inventory: Code execution via SDK calls (
generateContent), network operations, and package installation. - Sanitization: Absent; the skill does not specify any filtering or validation for the fetched documentation.
Audit Metadata