LayerFSv1.1.0 releases

Plugins

Plugins

Tools built on LayerFS. Each one pins its own commit and the LayerFS commit it was tested against, and carries a prompt your coding agent follows to install exactly those.

Submit a plugin
Tags

Build a plugin

A plugin builds on LayerFS’s JavaScript packages: @layerfs/agent for agents’ work, @layerfs/backend to show a project and @layerfs/runtime to run projects for users. Their wiki says which you need. Paste this into a coding agent in your plugin’s repository: it reads the wiki, designs around the packages and LayerFS’s rules, and waits for your OK before writing code. It builds on LayerFS v1.1.0 at 9f97a7ec95 and records that commit in layerfs.json, where the listing prompt finds it.

Build prompt for your agent
Build a LayerFS plugin with me: <what it should do>.

1. Clone https://github.com/Ephemeral-AI-Lab/layerfs beside this repository and check out commit 9f97a7ec95b662e9cb288704afa165feff132cfa. Build against that commit only, and record it in this repository as layerfs.json: { "commit": "<the 40-character SHA>" }. The listing prompt reads it from there.
2. Read api/README.md in that checkout: it maps LayerFS's JavaScript packages and links each one's guide. Decide which kind of plugin this is, and tell me before writing code:
   - a tool or skill for agents: @layerfs/agent. Agents that only have Bash use its layerfs command.
   - something that shows a project: also @layerfs/backend.
   - a host that runs projects for its users: also @layerfs/runtime.
3. Read the guides for the packages you chose, and the skill in .agents/skills/layerfs-js-sdk/. If agents will call LayerFS through the plugin, give them that skill.
4. Design within these rules:
   - Use only the packages' public calls. Never import @layerfs/agent/internal or LayerFS source files.
   - Settings are the LAYERFS_* environment names. If the host can only set its own names, map them in the plugin.
   - Store calls run in the calling process on Linux only. On macOS and Windows, start the worker with @layerfs/runtime and pass its endpoint to agents (LAYERFS_WORKER_ENDPOINT).
   - Each call is attempted once. Handle a refusal (Conflict, Busy, NoSuchSandbox, …) the way the guides say; never retry or replay a call on your own.
   - Change nothing in the LayerFS checkout. If the plugin needs something LayerFS doesn't have, stop and tell me.
5. Write a short design: which packages and calls, where each part runs, and what agents see. Wait for my OK.
6. Build it in this repository, with tests that run without Docker by default, and a README with an Install section a coding agent can follow.
7. When it works and is pushed, list it with the prompt at https://layerfs.ai/plugins/submit/.

If a step fails, stop and give me the exact error; don't retry it on your own.

How a plugin gets listed

  1. 01

    Build

    Write your plugin in its own repository, on LayerFS’s JavaScript packages; the prompt above has your agent do it. Your code stays there.

  2. 02

    Copy the template

    In your fork of LayerFS, copy docs/plugins/template/ to plugins/<your-plugin-id>/ and fill in its four files: plugin.json, the display page, the install prompt and the logo.

  3. 03

    Open a pull request

    Titled “Add plugin: <name>”, from your fork and under your name.

  4. 04

    Merged and listed

    We review it. Once merged, you are a LayerFS contributor and the plugin appears here.

Installing a plugin runs its author’s build on your machine. Read a plugin’s install prompt before you paste it.Official plugins are published by the LayerFS team; every other listing is reviewed but maintained by its author.