eval-chat-traces
Fail
Audited by Snyk on Apr 5, 2026
Risk Level: HIGH
Full Analysis
HIGH W007: Insecure credential handling detected in skill instructions.
- Insecure credential handling detected (high risk: 0.80). The prompt instructs the agent to pull and present raw chat traces (including the "Query" field) from logs without any redaction guidance, so if those traces contain API keys, tokens, or passwords the LLM would likely output them verbatim, creating an exfiltration risk.
Issues (1)
W007
HIGHInsecure credential handling detected in skill instructions.
Audit Metadata