vocs-documentation

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

• Third-party content exposure detected (medium risk: 0.30). Outsider-authored free text is ingested indirectly only when the required runtime exposes Vocs MCP endpoints (e.g., list_pages/read_page/search_docs or configured sources such as GitHub) that an external user/agent can query for arbitrary page/source contents, making poison possible if untrusted authors can publish into those sources.

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

• Potentially malicious external URL detected (high risk: 0.90). The docs include explicit agent-facing prompts that instruct an AI to "Read https://docs.my-sdk.dev/react/getting-started" and point agents at an MCP endpoint (https://docs.my-sdk.dev/api/mcp) plus a GitHub repo link (https://github.com/org/repo), which are runtime URLs an agent would fetch to control its instructions and to read/execute repository source via the MCP configuration.