AgentDB Memory Patterns

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions utilize npx agentdb@latest to fetch and execute the AgentDB CLI tool directly from the npm registry for database initialization and server management.
  • [COMMAND_EXECUTION]: The skill provides numerous CLI commands for initializing databases, starting MCP servers for Claude Code, and creating learning plugins, which involve local system operations.
  • [PROMPT_INJECTION]: The skill implements a memory storage and retrieval system that creates a surface for indirect prompt injection by ingesting and replaying untrusted data.
  • Ingestion points: The skill code demonstrates memory storage via insertPattern, storeMemory, and storeFact in SKILL.md (e.g., lines 63, 86, and 102).
  • Boundary markers: No specific delimiters or instructional guards are present in the provided examples to differentiate retrieved memory from current system instructions.
  • Capability inventory: The agent has access to CLI execution (npx) and plugin generation logic.
  • Sanitization: No input validation or content filtering is shown for data before it is persisted or after it is retrieved for agent reasoning.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 10:51 AM