Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
akunzai avatar

Mem Setup

  • 8 installs
  • Updated August 4, 2026
  • akunzai/agent-skills

mem-setup is a Claude Code skill that bridges each installed coding agent's native global memory file to a single canonical ~/.agents/AGENTS.md for shared cross-agent memory.

About

mem-setup wires each installed coding agent's native global memory file to a single canonical ~/.agents/AGENTS.md so Claude Code, Codex, OpenCode, Antigravity/Gemini, and pi all read the same durable cross-agent memory. It uses @import, symlink, or config instructions depending on each agent. A developer runs it to set up or refresh shared global memory across agents. The bundled script supports a dry-run plan before apply and backs up real files first.

  • Points every installed agent's global memory file at one canonical ~/.agents/AGENTS.md so Claude Code, Codex, OpenCode,
  • Bridges via @import, symlink, or config instructions[] depending on the agent's hook support
  • Ships a read-only plan/apply script with backups and idempotent re-runs

Mem Setup by the numbers

  • 8 all-time installs (skills.sh)
  • Ranked #12,339 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
  • Data as of Aug 5, 2026 (Skillselion catalog sync)
At a glance

mem-setup capabilities & compatibility

Capabilities
memory recall · memory promote · memory sync
Use cases
memory · orchestration
Platforms
macOS · Linux · Windows
Pricing
Free
From the docs

What mem-setup says it does

Points every installed agent's global memory at a single canonical core file so durable instructions are shared across agents.
SKILL.md
This skill replaces the retired `memory-autoload` plugin and its `~/.agents/MEMORY.md`.
SKILL.md
npx skills add https://github.com/akunzai/agent-skills --skill mem-setup

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs8
Last updatedAugust 4, 2026
Repositoryakunzai/agent-skills

What it does

Bridge every installed coding agent's global memory file to one canonical cross-agent AGENTS.md.

Who is it for?

Sharing one durable global memory file across Claude Code, Codex, OpenCode, Gemini, and pi

Skip if: Authoring the memory content itself, which the user must supply

When should I use this skill?

Setting up or refreshing global memory bridging across agents

What you get

Every installed agent points at one canonical ~/.agents/AGENTS.md so durable instructions are shared.

  • Bridged per-agent global memory files pointing at the canonical core
  • A read-only setup plan before applying

By the numbers

  • bridges 5 agents (Claude Code, Codex, pi, Antigravity/Gemini, OpenCode)

Files

SKILL.mdMarkdownGitHub ↗

mem-setup — Cross-Agent Global Memory Bridge

Points every installed agent's global memory at a single canonical core file so durable instructions are shared across agents. Ships the wiring only — you supply the content of ~/.agents/AGENTS.md (and any augmentation modules).

Canonical layout (user-owned)

  • ~/.agents/AGENTS.md — the cross-agent core, the single source of truth.
  • Optional augmentation modules — extra rules inlined for agents with weak

hook/skill support. Pass their paths via MEM_SETUP_AUGMENT (colon-separated). The repo ships none.

Per-agent bridge

AgentNative global pathTierMethod
Claude Code~/.claude/CLAUDE.mdhigh (core)@import
Codex~/.codex/AGENTS.mdhigh (core)symlink (copy on Windows)
pi~/.pi/agent/AGENTS.mdhigh (core)symlink (copy on Windows)
Antigravity / Gemini~/.gemini/GEMINI.mdlow (core + augment)@import
OpenCode~/.config/opencode/opencode.jsonlow (core + augment)instructions[]

High-tier agents get dynamic behaviors from plugins, so they read the core only. Low-tier agents have weak hook support, so augmentation modules are composed into their static prompt via import / config — never by duplicating a file.

Workflow (always dry-run first)

1. Ensure ~/.agents/AGENTS.md exists (your durable cross-agent instructions). 2. Run the read-only plan and show it to the user: bash skills/mem-setup/scripts/mem-setup-bridge.sh plan 3. After explicit user confirmation, apply: bash skills/mem-setup/scripts/mem-setup-bridge.sh apply

Override the core path with MEM_SETUP_CANONICAL and augmentation modules with MEM_SETUP_AUGMENT (colon-separated absolute paths).

Safety

  • Only acts on installed agents (those whose config directory exists).
  • Never write through a symlink — an existing symlink is removed first, so the

canonical core is never clobbered.

  • Real files are backed up to *.bak-<timestamp> before any change.
  • Idempotent — re-running makes no changes once bridged.
  • Windows (Git Bash): import/config methods need no symlink; for Codex/pi the

script falls back to a copy and warns that you must re-run to refresh.

Migration from memory-autoload

This skill replaces the retired memory-autoload plugin and its ~/.agents/MEMORY.md. If that plugin is still installed, remove it; if ~/.agents/MEMORY.md holds content, merge it into ~/.agents/AGENTS.md first.

Related skills

FAQ

Which agents does mem-setup bridge?

Claude Code, Codex, pi, Antigravity/Gemini, and OpenCode, each via @import, symlink, or config instructions.

Is it safe to re-run?

Yes, it is idempotent, only acts on installed agents, backs up real files, and never writes through a symlink.

AI & Agent Buildingagentsautomation

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.