audit-context-building

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill mandates ultra-granular, line-by-line, evidence-based code analysis with line citations and literal block quoting (for continuity and anti-hallucination), which effectively requires reproducing exact code strings—so if secrets are present in the code or external-call payloads, the LLM would be expected to output them verbatim, creating exfiltration risk.