reflect

Warn

Audited by Snyk on Aug 20, 2026

Risk Level: MEDIUM
Full Analysis

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

  • Third-party content exposure detected (medium risk: 0.30). The reflect workflow ingests the active session transcript (which may include outsider-provided free text) at the reviewers’ step (“Read the active transcript at <ABSOLUTE_PATH>”), without first selecting a specific trusted item, so prompt-injection text can enter the LLM via the transcript.

Issues (1)

W011
MEDIUM

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

Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 20, 2026, 03:21 PM
Issues
1
Security Audit — snyk — reflect