rombot-community

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

• Third-party content exposure detected (high risk: 0.75). The required workflow sends the operator’s question over HTTPS to RomBot and then ingests RomBot’s JSON field answer (outsider-authored community response text fetched at runtime) from scripts/rombot-ask.py response parsing (payload = json.loads(...); answer = payload.get("answer")) and prints it to stdout, with SKILL.md requiring it be passed through “verbatim” into the agent/human context.

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

• Potentially malicious external URL detected (high risk: 0.90). The CLI reads ROMBOT_ASK_URL at runtime and POSTs the question to the configured endpoint (example URL "https://api.rombot.uk/api/community-ask"), then prints RomBot's returned "answer" verbatim (i.e., the remote response directly controls the agent's output), so the remote URL controls prompts at runtime.