bkend-storage

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to "use WebFetch" to retrieve live documentation from public raw.githubusercontent.com URLs (e.g., the Storage Overview and MCP Storage Guide links), which are open, third‑party, user‑authored pages the agent would read and could let external content influence its behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly tells the agent to use WebFetch to retrieve live docs from raw.githubusercontent.com (e.g. https://raw.githubusercontent.com/popup-studio-ai/bkend-docs/main/en/storage/01-overview.md and the other raw.githubusercontent.com links), which would be fetched at runtime and injected into the agent context to directly drive generated prompts/code.