sequential-thinking

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt requires citing "file:line" evidence for every claim and making tasks for each file read, which effectively forces the agent to read and reproduce file contents (potentially including API keys/passwords) verbatim — creating an exfiltration risk.