humanize

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.95). This skill requires reproducing code, quotes, numbers, names and other tokens verbatim when humanizing user-provided text, so if the input contains API keys, passwords, or other secrets the model is instructed to output them exactly, creating a high exfiltration risk.