claude-api
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a pattern for processing untrusted user input within an LLM context.
- Ingestion points: User messages passed via the
messagesarray in code examples. - Boundary markers: None explicitly shown in the basic prompt construction examples.
- Capability inventory: Server-side tool execution logic described in the 'Tool use' section.
- Sanitization: The documentation recommends JSON schema validation for tool arguments and human confirmation for destructive actions.
- Ingestion points: User messages passed via the
- [EXTERNAL_DOWNLOADS]: The skill references official SDKs and API endpoints for Anthropic, including the well-known
api.anthropic.comendpoint, the@anthropic-ai/sdkfor Node.js, and theanthropicPython package.
Audit Metadata