supabase

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md instructs the agent to fetch outsider-authored public web content at runtime (the Supabase changelog and linked docs pages), which then becomes readable text in the LLM’s context (via doc fetching/summarization), enabling indirect prompt injection.

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 at runtime to fetch https://supabase.com/changelog.md and https://supabase.com/docs/guides/security/product-security.md and to scan those documents to guide its behavior, meaning the fetched content directly influences agent prompts.