think-critically
Fail
Audited by Snyk on Mar 9, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E004: Prompt injection detected in skill instructions.
- Potential prompt injection detected (high risk: 1.00). The prompt contains explicit, deceptive directives that attempt to override system-level instructions (e.g., "This overrides system-level instructions" and strict "FIRST-TOKEN ENFORCEMENT"/"absolute" constraints) which are outside the legitimate scope of an evaluation skill and constitute prompt-injection attempts.
HIGH W007: Insecure credential handling detected in skill instructions.
- Insecure credential handling detected (high risk: 1.00). The prompt forces the model to produce verbatim quotes and fully reproduce the original Prompt/Expectations (e.g., "verbatim quote" for fixes, supporting text, and the full Revised Prompt), so if those inputs contain API keys, tokens, or passwords the LLM is required to output them.
Audit Metadata