ai-wrapper-product
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references the official Anthropic SDK for Node.js.
- Evidence: The code snippet imports @anthropic-ai/sdk to facilitate communication with AI models (SKILL.md).
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for interpolating external user data into prompt templates, which constitutes an attack surface for indirect prompt injection.
- Ingestion points: User-provided content is ingested via the userInput parameter in the generateContent function and through fields in the promptTemplates object (SKILL.md).
- Boundary markers: While the examples structure inputs as labeled fields, they lack explicit instructions or delimiters designed to prevent the model from interpreting user data as new instructions (SKILL.md).
- Capability inventory: The skill demonstrates executing requests to the Anthropic API via the @anthropic-ai/sdk library (SKILL.md).
- Sanitization: The provided code includes a basic length check for user input and the architecture diagram references validation and sanitization processes (SKILL.md).
Audit Metadata