check-known-solutions

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow calls the Push Realm MCP server (search_solutions) and then ingests the returned “matches/next_action/hint” text into the agent’s LLM context, where that text is authored by an external community (outsider source: public/shared knowledge base content fetched at runtime via MCP).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls the Push Realm MCP endpoints (https://api.pushrealm.com/mcp) at runtime (search_solutions, submit_solution, record_agent_usage) to fetch community-provided fixes that the agent is instructed to apply, so external content can directly control agent instructions or suggest code/commands to execute.