AgentDB Memory Patterns
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the
agentdbpackage from the NPM registry vianpx agentdb@latestto perform various administrative and operational tasks. - [COMMAND_EXECUTION]: Provides instructions to execute shell commands for initializing databases, managing learning plugins, and starting an MCP server for integration with Claude Code.
- [DATA_EXFILTRATION]: Accesses local database files and filesystem paths (e.g.,
./agents.db,.agentdb/reasoningbank.db) to store and retrieve agent memory patterns. It also includes functions to export vector data to local JSON files. - [REMOTE_CODE_EXECUTION]: Uses
npxto execute remote code from the AgentDB package at runtime. This is a standard method for utilizing the AgentDB CLI tools as documented by the vendor. - [DYNAMIC_EXECUTION]: Includes a plugin creation system (
create-plugin) that generates code from templates (e.g., decision-transformer, q-learning) for agent logic, which is a core feature of the skill's learning architecture.
Audit Metadata