claude-api

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill explicitly enables and instructs use of WebFetch/WebSearch and live documentation URLs (see shared/live-sources.md and the SKILL.md / python/agent-sdk/README.md examples such as "Open example.com and describe what you see"), so the agent is expected to fetch and interpret public third‑party webpages (platform.claude.com, GitHub, arbitrary sites) which could contain untrusted/user-generated instructions that influence tool use and subsequent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs using WebFetch at runtime to fetch live docs (e.g., https://raw.githubusercontent.com/anthropics/claude-agent-sdk-python/main/README.md) and uses the fetched content with extraction prompts to control model instructions, so this URL is used at runtime and its content directly controls prompts.