Agent Memory Hub
ReviewAudited by ClawScan on May 16, 2026.
Overview
Review recommended: this is a coherent local memory tool with no exfiltration shown, but it can save secrets and personal details in persistent plaintext memory that future agents may reuse.
Install only if you are comfortable with a local, persistent memory file. Configure the storage directory if needed, periodically inspect or delete `~/.agent-memory/memories.json`, and tell your agent not to store passwords, API keys, tokens, or other sensitive information.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Passwords, API keys, client details, or personal facts shared with an agent could be saved across sessions and returned later or used to influence future answers.
The skill explicitly supports credential-related memories and stores persistent memory as readable local JSON, creating a risk that secrets or sensitive personal data are retained and later surfaced to agents.
Good keys: `user_name`, `preferred_stack`, `project_deadline`, `api_key_note` ... Storage: `~/.agent-memory/memories.json` ... Format: human-readable JSON ... Auto-Tag Categories ... `credential`
Do not store secrets in this memory. The publisher should add secret redaction or denylisting, explicit user approval for sensitive memories, retention controls, and clear instructions for inspecting and deleting the memory file.
Users have less provenance information for deciding whether the local Node MCP server and its dependency chain are trustworthy.
The artifacts include source and package files, but the registry metadata does not provide a verifiable source location or install spec, so users must rely on the supplied package contents and npm build process.
Source: unknown; Homepage: none; Install specifications: No install spec — this is an instruction-only skill.
Verify the package contents before building, use the included lockfile, and prefer a published source repository or signed release if available.