subagent-driven-development

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs subagents to "use verbatim" exact values copied from plan/global-constraints and to place "Exact values (numbers, magic strings, signatures...)" in briefs, which would force an LLM-driven implementer/reviewer to reproduce those literal strings (including any secrets) in generated code/commands — creating an exfiltration risk.