ultra-prompt

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt mandates auditing configs, logs, and other repo files and to "use concrete repository evidence" when populating the output template but gives no instruction to redact or avoid embedding secret values, so the agent may be forced to read and output API keys, tokens, or passwords verbatim.