darwin-dynamic-agentically-rewriting

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the LLM to receive full source-code chunks and return modified code (and diffs/memory entries), so if those inputs contain API keys, tokens, or passwords the model will be asked to reproduce them verbatim in its outputs — creating a high exfiltration risk.