agent-memory

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: Analysis of the Python and TypeScript code for file-based and database-backed memory found no evidence of malicious intent, obfuscation, or unauthorized data exfiltration. All file and network operations are consistent with the stated purpose of building a persistent knowledge base.
  • [PROMPT_INJECTION]: The skill facilitates retrieval-augmented generation (RAG), which creates an inherent surface for indirect prompt injection if stored memories contain malicious instructions.
  • Ingestion points: Files in the 'memory' directory, SQLite database records, and ChromaDB collections.
  • Boundary markers: Not implemented in the provided snippets; the documentation suggests injecting memories into system prompts without recommending specific delimiters.
  • Capability inventory: Local file system access (read/write), SQLite database operations, and network calls to OpenAI's embedding API.
  • Sanitization: None implemented in the templates; content filtering is left to the developer implementing the memory system.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 08:15 AM