memory-management

Pass

Audited by Gen Agent Trust Hub on Apr 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Potential for Indirect Prompt Injection surface. The skill ingests user input and stores it in profile.md, habit.md, and context.md files, which are later read back using read_file or deep_recall_memory to guide agent behavior.
  • Ingestion points: Data entering through user dialogue is saved to the .agent_memories/ directory (SKILL.md).
  • Boundary markers: Absent. There are no instructions to use delimiters or warnings to ignore embedded instructions when reading stored content.
  • Capability inventory: The skill utilizes read_file, edit_file, and deep_recall_memory capabilities (SKILL.md).
  • Sanitization: Minimal. While SKILL.md instructs the agent not to record sensitive data, there is no technical sanitization or validation of the input content to prevent malicious instruction embedding.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 13, 2026, 02:46 AM
Security Audit — agent-trust-hub — memory-management