The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones a repository from an external source at https://github.com/webzler/agentMemory.git. This repository is outside the trusted environment and its contents are not verified.
[REMOTE_CODE_EXECUTION]: Following the download, the skill executes npm ci and npm run compile on the third-party code. This action can trigger arbitrary execution via package lifecycle scripts or compiled binaries bundled within the repository.
[COMMAND_EXECUTION]: The setup and operational workflow utilizes multiple shell commands, including git clone, git checkout, and npm run start-server, allowing the agent to perform network and system-level operations.
[PROMPT_INJECTION]: The skill provides tools for data ingestion (memory_write) and retrieval (memory_read, memory_search), creating an attack surface for indirect prompt injection. Malicious instructions stored in the memory bank could influence the agent's logic when retrieved in future contexts.
Ingestion points: memory_write tool in SKILL.md.
Boundary markers: None present to delimit stored memories from active instructions.
Capability inventory: Subprocess calls via npm run, file system access for memory storage, and network operations for MCP communication.
Sanitization: No explicit sanitization or filtering of stored content is described.

agent-memory-mcp