learn-from-correction

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt requires the agent to "always include the source correction" and verbatim-quote agent and human outputs in the proposal, which forces the LLM to reproduce any secrets (API keys, tokens, passwords, cookies) present in those outputs and therefore creates an exfiltration risk.