nightly-eval-investigation
Warn
Audited by Snyk on Nov 15, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The required workflow downloads nightly eval artifacts (trajectories, raw agent session logs, and
evals.jsonpage snapshots) from the internal GCS bucketgs://guidance-evals/, and outsiders cannot post into that bucket, but these logs may contain web content the eval agents fetched during their own runs, so the agent must read outsider-influenced text only after selecting these artifacts.
Issues (1)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
Audit Metadata