recipe-generalization-sweep

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The recipe explicitly requires recording and emitting verbatim quotes from files as evidence (e.g., "file:line + verbatim quote" and including those quotes in the written solution_outline), which forces the LLM to read and output exact repository text and could exfiltrate any secrets present there.