blog

Fail

Audited by Snyk on Feb 20, 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 requires preserving code blocks and frontmatter exactly and to use the source content verbatim, so if those contain API keys/passwords the LLM must reproduce them in output, creating an exfiltration risk.
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 20, 2026, 08:39 AM