agent-memory
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill stores and retrieves arbitrary user content, which introduces an indirect prompt injection surface. Malicious instructions stored in memory could influence the agent when recalled in future sessions. 1. Ingestion points: User-provided text ingested via the 'memory store' command. 2. Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands when recalling memories. 3. Capability inventory: The skill manages a local database and provides retrieval tools. 4. Sanitization: No evidence of input sanitization or filtering was found.
- [COMMAND_EXECUTION]: The 'scripts/memory-init.sh' script initializes the database and schema using sqlite3. It correctly applies restrictive file permissions (chmod 600 and 700) to the storage directory and database files, following security best practices.
- [EXTERNAL_DOWNLOADS]: The documentation provides instructions to clone the skill from the vendor's official GitHub repository (github.com/G-HunterAi/agent-memory.git) for installation.
Audit Metadata