context-engineering-collection

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

• Third-party content exposure detected (medium risk: 0.30). This repo’s “researcher” continuous loop runs a required runtime workflow that can ingest queue/inbox text items at runtime (via researcher/scripts/loop_step.py reaping/pulling from researcher/queue/*.jsonl), and those queue entries can plausibly be attacker-authored free text (e.g., if an outsider can post/submit work items into the workflow’s feed/queue), enabling indirect prompt injection.