remember

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFE
Full Analysis
  • Persistent Memory Storage: The skill utilizes file-writing tools to update AGENTS.md (both globally at ~/.deepagents/agent/AGENTS.md and within project directories). This allows the agent to maintain user preferences and project-specific guidelines across different conversation sessions.
  • Dynamic Skill Creation: The instructions guide the agent in creating new, reusable skill modules within the ~/.deepagents/agent/skills/ directory. This process includes generating a directory structure for documentation and executable scripts, which is an intended mechanism for automating recurring workflows discovered during conversation.
  • Data Ingestion Surface: To identify learnings, the skill scans the current conversation history. Because this involves processing dialogue that may contain information from external or untrusted sources, it presents a standard surface for indirect prompt injection where instructions could potentially be persisted into memory. This is a common consideration for all agentic memory features.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 01:20 AM