cross-agent-delegation

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs the agent to embed the full mnemos "CONTEXT" and other query outputs verbatim into shell commands/prompts (e.g., the kimi -p string), which would force any secret values present in those checkpoints or query outputs to be included in the LLM's generated output and passed on—an exfiltration risk.