observability-designer

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

• Third-party content exposure detected (medium risk: 0.30). At runtime the workflow reads outsider-authored JSON alert/service definitions from a user-provided --input file (e.g., scripts/alert_optimizer.py --input ... then load_alert_config() via json.load() and processes its free-text fields like expr, alert/name, and annotations to generate analysis/report output), meaning an outsider can supply arbitrary strings for the LLM to ingest indirectly through the JSON contents.