memory-system
Pass
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Automatically downloads the all-MiniLM-L6-v2 model (~80MB) from Hugging Face via the sentence-transformers library. Hugging Face is a well-known and trusted service.
- [COMMAND_EXECUTION]: Executes indexing, searching, and maintenance tasks via Python CLI commands in scripts/memory.py, which uses parameterized SQL to interact with a local database safely.
- [PROMPT_INJECTION]: Handles untrusted data from local Markdown files, creating a surface for indirect prompt injection. 1. Ingestion points: Markdown files in the memory directory. 2. Boundary markers: SKILL.md provides instructions to the agent to distinguish between search context and instructions. 3. Capability inventory: File system access (read/write in memory/) and SQLite operations. 4. Sanitization: SQL parameterization is used throughout the script.
Audit Metadata