Skills · last verified 2026-10-24

thedotmack/claude-mem — install for Claude Code

github.com/thedotmack/claude-mem · persistent memory layer

What it does

A persistent memory layer for coding agents: it remembers context across sessions, so the agent picks up where it left off instead of starting from zero each time.

Memory skills are the trust-wedge flagship — the whole job is storing what the agent learns, which makes the injection surface the point of the product, not an afterthought.

Install

Generic / all IDEs (single command, verified against the repo README's Quick Start):

npx claude-mem install

Claude Code (plugin marketplace):

/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

Per-IDE variants: npx claude-mem install --ide opencode (also --ide t3code, --ide antigravity, --ide omp, --ide pi, --ide dsh --dsh-profile tui, --ide grok-bot).

OpenClaw Gateway: curl -fsSL https://install.cmem.ai/openclaw.sh | bash

Note: the installer now asks you to sign in (an email link, no card required) and uses hosted memory by default, with a 30-day free trial; you can skip sign-in with an explicit provider choice, and a local observer is opt-in.

Verified against github.com/thedotmack/claude-mem (README, Quick Start), 2026-10-24. No Cursor-specific command is documented — the generic npx claude-mem install is the documented path there.

Trust score: 5/10

• Skill-file review (2/3): a memory layer holds exactly what you tell it — and what the agent tells it. Review what's stored and where the store lives before it accumulates real project context.
• Install behavior (2/3): the npx CLI runs install code on your machine; the OpenClaw path is a curl-piped-to-shell install — the highest-privilege install pattern in this series.
• Prompt-injection surface (1/4): this is the highest injection surface of the five pages: persistent memory means untrusted content seen once can steer the agent weeks later. That is the product's whole point, which is why the bar here is higher than elsewhere.

How the score works: skill-file review, install behavior, and prompt-injection surface, each scored against what the skill asks for beyond its stated job. This is a public-code review as of 2026-10-24 — a signal, not a certification. Full methodology: skills index.

Also in this series

• obra/superpowers — install for Claude CodeTrust 6/10 · 2026-10-24
• mattpocock/skills — install for CursorTrust 6/10 · 2026-10-24
• Leonxlnx/taste-skill — install for Claude CodeTrust 6/10 · 2026-10-24
• upstash/context7 — install for Claude CodeTrust 5/10 · 2026-10-24