mem0-memory

Warn

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The script scripts/visualization/setup-visualization-deps.sh dynamically installs several Python packages, including plotly, networkx, and matplotlib, using pip. Runtime installation of dependencies from external registries is a security risk if the registry or the packages themselves are compromised.
  • CREDENTIALS_UNSAFE (MEDIUM): Multiple scripts, such as scripts/crud/add-memory.py, scripts/crud/search-memories.py, and scripts/crud/delete-memory.py, accept the MEM0_API_KEY via a command-line argument (--api-key). This practice exposes sensitive credentials to other users on the same system via process monitoring tools like ps and saves the secrets in the user's shell command history.
  • PROMPT_INJECTION (MEDIUM): The skill is vulnerable to Indirect Prompt Injection (Category 8). Ingestion points: scripts/create/create-all-agent-memories.py reads content from all .md files in the agents/ directory. Boundary markers: None identified; content is interpolated directly into memory messages. Capability inventory: The skill uses client.add() and client.update() to write data to the remote Mem0 memory store via network calls. Sanitization: No validation, filtering, or escaping of the ingested markdown content is performed before storage.
  • COMMAND_EXECUTION (LOW): The script scripts/visualization/quick-visualize.sh uses the open command on macOS to launch a generated dashboard in the default browser.
  • DATA_EXPOSURE (LOW): The client initialization library in scripts/lib/mem0_client.py attempts to load configuration from ~/.mem0.env in the user's home directory. Accessing files outside the project scope increases the attack surface for local sensitive data exposure.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 15, 2026, 09:47 PM