claude-api
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a library of official documentation and best-practice templates for building Claude-powered applications. It covers essential topics like streaming, tool use, batch processing, and agentic workflows.
- [EXTERNAL_DOWNLOADS]: Fetches latest API specifications and package metadata from official repositories and well-known services (Anthropic, Google, MCP, npm, PyPI). These references are documented neutrally and follow standard development practices.
- [COMMAND_EXECUTION]: Provides implementation guidance for tools involving shell execution (Bash) and browser automation (Playwright MCP). The documentation explicitly recommends using permission modes (e.g., 'default' prompt) and sandboxed environments (Anthropic's hosted code execution) to mitigate risks.
- [DATA_EXFILTRATION]: Includes proactive security warnings against hardcoding API keys and suggests secure storage methods like environment variables. It also advises on protecting PII and sensitive data when using persistence features like the memory tool.
- [DYNAMIC_EXECUTION]: Explains how to use the official server-side code execution tool, which runs in an isolated, sandboxed container on Anthropic's infrastructure, ensuring local environment safety.
Audit Metadata