AgentDB Memory Patterns

This skill description provides coherent APIs and CLI commands for persistent agent memory using AgentDB. The content itself doesn't include explicit malicious code, hard-coded credentials, or obvious exfiltration endpoints. However, it relies heavily on npx-based installs and plugin/training features that could execute arbitrary remote code and access local DB files. That introduces a supply-chain risk: unpinned npx installs and plugin-driven transitive installs could result in code that reads or transmits memory data. Recommend pinning package versions, auditing the agentdb/agentic-flow package source, reviewing plugin code before use, and treating imported JSON data as untrusted. Overall: not confirmed malware, but moderate supply-chain/security risk requiring review.