deep-agents-memory

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • Filesystem Interactions: The skill provides tools for the agent to interact with the local filesystem. These are configured with security features such as virtual_mode to prevent unauthorized directory access and interrupt_on to require user approval for file changes.
  • State Management and Persistence: The architectural patterns for memory backends described in the skill are standard for maintaining agent state across sessions and do not involve unauthorized data movement.
  • Data Ingestion Surfaces: Since the skill allows the agent to read local files, it creates a surface for indirect instructions within that data. This is a common aspect of filesystem-enabled agents, and the skill provides the framework for managing this capability.
  • Secure Configuration Practices: The skill uses placeholders for connection details in its production storage examples, following the best practice of avoiding hardcoded credentials in documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 05:42 PM
Security Audit — agent-trust-hub — deep-agents-memory