claude-docs

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

• Third-party content exposure detected (high risk: 0.75). The required workflow instructs the agent to “WebFetch the canonical URL” from platform.claude.com (public web content) and then use that fetched page text as LLM context, which is an outsider-authored free-text source at runtime.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs the agent to "WebFetch" the canonical documentation at https://platform.claude.com/docs/en/ (and linked pages) at runtime, which means external content from that URL is fetched and injected into the agent context to control prompts/responses.