agent-memory-mcp
Warn
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from an external source (https://github.com/webzler/agentMemory.git) that is not part of the established trusted organizations.
- [COMMAND_EXECUTION]: The setup process involves executing multiple shell commands, including 'npm install', 'npm run compile', and starting a local MCP server.
- [REMOTE_CODE_EXECUTION]: By downloading an external repository and immediately running its installation and compilation scripts, the skill creates a path for potentially malicious code to be executed on the local system.
- [PROMPT_INJECTION]: The skill implements a memory system that ingests and searches untrusted data through tools like 'memory_write' and 'memory_search', creating a surface for indirect prompt injection. Ingestion points: External data enters the agent context via 'memory_write' in SKILL.md. Boundary markers: No delimiters or instructions are provided to delimit or ignore embedded instructions. Capability inventory: The skill allows shell command execution (npm install, compile) and local server operations. Sanitization: There is no evidence of sanitization or validation of content processed by the memory tools.
Audit Metadata