ai-model-cascade
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides implementation patterns and code examples that are vulnerable to indirect prompt injection attacks where untrusted data could be interpreted as instructions by the AI model.
- Ingestion points: The
generateWithFallbackfunction and its usage examples (such as inExample 1,Example 2, andExample 5inSKILL.md) demonstrate prompt construction via direct string interpolation of variables liketranscript,question, andtext. These variables often contain external, untrusted content. - Boundary markers: The prompt templates provided in the documentation do not use delimiters (e.g., XML tags, triple backticks) or "ignore embedded instructions" warnings to isolate user-provided data from system instructions.
- Capability inventory: The skill facilitates network communication with the Google Gemini API and produces structured JSON output intended to be parsed and used by downstream application logic.
- Sanitization: The implementation lacks any mechanisms for sanitizing, escaping, or validating the input data before it is interpolated into the prompts.
Audit Metadata