agent-memory-mcp

The manifest describes a legitimate-sounding memory/indexing skill, but the installation/execution model presents a material supply-chain and data-exposure risk. The primary danger is executing unreviewed third-party JavaScript (npm install + start scripts) with broad filesystem access to an absolute workspace path. Without reviewing the actual repository and its dependencies, one cannot rule out network exfiltration or malicious lifecycle scripts. Treat this as SUSPICIOUS: safe to use only after manual code and dependency review, running inside a sandbox with strict network/FS controls.