humanizing-text

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly tells the agent to "preserve unaffected spans and exact material such as quotations, code, commands, links, identifiers, and data," which requires verbatim reproduction of any sensitive tokens or credentials present in the input, creating a high exfiltration risk.