oss-analyze

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

• Third-party content exposure detected (high risk: 0.95). Outsider-authored free text is ingested when the skill fetches and reads an arbitrary GitHub repository at runtime (e.g., via gitmcp.io MCP get_file_content / get_file_tree or curl raw.githubusercontent.com / git clone), then parses README/docs/source files into the generated knowledge/oss-feature-matrix.md that is fed back into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches repository content at runtime (e.g., https://gitmcp.io/{owner}/{repo}, raw.githubusercontent.com raw file URLs, or via gh repo clone / git clone) and injects that remote content into the agent's context to drive the generated feature matrix, so the fetched files can directly control the agent's prompts/output.