team-adversarial-swarm

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow calls wf-swarm-explore.js, wf-swarm-score.js, wf-swarm-converge.js, and wf-swarm-synthesize.js, and these agents ingest objective, rubric, historyDigest, and especially ant.evidence[].source / ant.evidence[].finding / candidate_solution.summary that originate from outsider-authored free text (e.g., user-provided task intent and any evidence strings produced by other agents or imported artifacts), which is then embedded directly into the LLM prompts via template literals.