nexus-room

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md describes runtime intake from the internal kanban #nexus channel via command/history reads (e.g., kanban channel history nexus -n 50) and exchanging “chunky summary updates,” which is outsider-submitted free text if external users can post into that channel.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs agents to call the wiki API at https://nexus.langwatch.ai (using NEXUS_PROD_URL + NEXUS_PROD_AGENT_TOKEN) and to "Read the writing-rules wiki page in full before drafting," which means runtime-fetched content from that URL will directly control the agent's instructions.