jobspipe-mcp

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md describes an MCP server whose tools return “live data directly” over Streamable HTTP (public web content fetched at runtime via the configured MCP endpoint), which can include outsider-authored free text that the agent may ingest into its 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 registers and fetches tool definitions from the MCP endpoints (e.g., https://jobspipe.dev/mcp and https://docs.jobspipe.dev/mcp, plus discovery manifests at https://jobspipe.dev/.well-known/mcp.json and https://jobspipe.dev/.well-known/mcp/server-card.json) at runtime, and those remote endpoints supply the agent's tools/specs which directly control agent behavior.