AgentDB Memory Patterns

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the agentdb package and associated plugin templates from the npm registry using npx.
  • [COMMAND_EXECUTION]: Provides CLI commands for database initialization, querying, and integration with the Model Context Protocol (MCP).
  • [DATA_EXFILTRATION]: Performs operations on local file paths like ./agents.db to manage vector storage and session history.
  • [PROMPT_INJECTION]: Identifies a surface for indirect prompt injection.
  • Ingestion points: Interaction memory storage and database query operations.
  • Boundary markers: Uses structured JSON for patterns, though specific delimiters are not demonstrated.
  • Capability inventory: Local file system access and plugin-based code execution.
  • Sanitization: Not explicitly documented in the provided examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 07:39 PM