Memory

claude-plugins

Public Claude Code and Codex plugins for local-first AI workflows.

The first plugin in this repo is memory: a setup skill for creating a local, git-backed Obsidian vault that captures durable notes from AI working sessions.

Read the companion post: Memory as reliability practice

Why This Exists

AI sessions can be useful in the moment and still disappear as soon as the thread ends. The assistant helped debug a strange failure, compare trade-offs, draft a review, or clarify a decision, but the useful context stayed trapped in a transcript.

memory is a small attempt to make that context durable without making it mysterious:

• Obsidian for plain Markdown that stays inspectable
• git for auditable history
• explicit setup gates before anything writes to dotfiles
• Claude Code session capture as the first-class path
• compatibility notes for Codex and Antigravity so the memory belongs to the person, not one assistant

The goal is not a grand second-brain system. The goal is lower re-orientation cost: fewer repeated explanations, fewer lost decisions, and a clearer trail from “we figured this out” to “we can reuse this later.”

What the memory Plugin Does

Skill	Purpose
/memory:memory-setup	Set up an Obsidian-backed memory vault with Claude Code session capture and optional Codex/Antigravity importers.

The default Claude Code path sets up:

• a local Obsidian vault under a user-selected parent directory
• claudian and obsidian-git as Obsidian community plugins
• a SessionEnd hook in ~/.claude/settings.json
• a managed memory block in ~/.claude/CLAUDE.md

Current support note: the setup flow has been tested on macOS. Linux and Windows compatibility reports and PRs are very welcome.

Install

Claude Code marketplace metadata lives in .claude-plugin/marketplace.json.

Codex marketplace metadata lives in .agents/plugins/marketplace.json.

For Codex:

codex plugin marketplace add adzuci/claude-plugins
codex plugin add memory@adzuci-plugins

For Claude Code:

claude plugin marketplace add adzuci/claude-plugins
claude plugin install memory@adzuci-plugins

Then invoke the setup skill:

/memory:memory-setup

Safety Model

This skill is intentionally cautious because it touches personal knowledge and local configuration.

• It explains the trade-offs before setup.
• It runs read-only preflight checks first.
• It asks before installing Obsidian.
• It backs up settings before editing.
• It keeps generated memory in plain Markdown.
• It is designed to be rerunnable and idempotent where possible.

The repo includes CI for the bundled Python scripts and plugin metadata:

• compile all memory setup scripts and tests
• run the pytest suite on Python 3.9 and 3.12
• validate marketplace and plugin JSON
• verify the static site references existing assets

Repository Structure

.claude-plugin/
  marketplace.json
.agents/
  plugins/
    marketplace.json
.github/
  workflows/
    ci.yml
docs/
  index.html
  assets/
plugins/
  memory/
    .claude-plugin/plugin.json
    .codex-plugin/plugin.json
    README.md
    skills/memory-setup/
      SKILL.md
      scripts/
      templates/
      references/
      tests/

Development

This repo is intentionally small, but the bar for changes is still: can someone else install it, understand what it writes, and recover cleanly if something goes wrong?

Run the memory skill tests:

python3 -m pytest plugins/memory/skills/memory-setup/tests -q

Validate the plugin metadata with the CLI available in your environment:

claude plugin validate .

Contributing

If this workflow is useful, please open an issue or PR. The most valuable feedback is concrete:

• setup failures on a real machine
• Linux or Windows compatibility fixes
• missing preflight checks
• unclear install instructions
• unsafe or surprising writes
• compatibility fixes for Claude Code, Codex, Antigravity, Obsidian, or git

Small, focused PRs are welcome. Please include the command you ran, the environment you tested on, and whether python3 -m pytest plugins/memory/skills/memory-setup/tests -q passed.

License

MIT