cross-agent-delegation

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs Claude to capture and embed full mnemos/icpg outputs (context, constraints, recent files, git state) verbatim into delegated agent prompts (kimi -p / mnemos resume), which can contain code/config secrets and forces the LLM to handle/output those secret values across agents.