batch-report

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

• Third-party content exposure detected (high risk: 0.90). The skill reads arbitrary batch output files (e.g., results.jsonl, output.jsonl, final_report.json, failed_urls.txt) from user-specified output directories and in batch_report/analysis.py (evaluate_llm_gates) it samples those file contents and sends them to a local LLM subprocess (SCILLM_SCRIPT) whose outputs are used as quality-gate metrics that directly affect decisions and reports, meaning untrusted third-party content can influence tool behavior.