memory-management

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt directs the agent to store and then inject raw "Context from memory: {context}" (including full user inputs and past interactions) directly into prompts, which means any secrets users enter could be fed to and reproduced by the LLM verbatim, enabling exfiltration.