agents
Use when designing, deploying, or debugging a Butterbase Agent (declarative LLM/tool graph), registering an MCP server for tool use, or wiring access controls…
Use when publishing a Butterbase app as a public template (visibility + listed + repo snapshot), browsing the template gallery, or cloning a template with environment-variable preflight. Templates are public apps with a pushed repo snapshot that other users can fork into their
$ npx -y skills add butterbase-ai/butterbase-skills --skill templates --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/templatesContext preview
The summary Claude sees to decide when to auto-load this skill.
Use when publishing a Butterbase app as a public template (visibility + listed + repo snapshot), browsing the template gallery, or cloning a template with environment-variable preflight. Templates are public apps with a pushed repo snapshot that other users can fork into their
name: templates description: Use when publishing a Butterbase app as a public template (visibility + listed + repo snapshot), browsing the template gallery, or cloning a template with environment-variable preflight. Templates are public apps with a pushed repo snapshot that other users can fork into their own region.
A "template" is a Butterbase app marked `visibility: 'public'` and `listed: true`, **with a repo snapshot uploaded via `butterbase repo push`**. The clone pipeline replays the source app's schema, RLS, functions, config, and frontend artifact into a new app in the cloner's region; the pushed repo gives the cloner the actual source tree to keep developing.
The repo snapshot is the **source-code half** of a template; the clone job replays schema/RLS/functions/config separately. Both halves are pinned via `apps.repo_latest_snapshot`.
**Reading a template's source code:** clone it into a scratch directory — `butterbase clone <source_app_id> /tmp/peek` — and read the resulting tree. There is no separate "browse template source" CLI; snapshots are only materialised by `repo pull`, which requires the caller to be bound to the app.
**Updating a clone with newer template versions:** clones don't auto-follow the source — each clone owns its own app and its own snapshot history. `butterbase repo pull` in the clone fetches the cloner's *own* latest, not the upstream publisher's. To track upstream changes, the cloner has to clone again into a sibling directory and merge with git, or the publisher has to push to a snapshot the clone can read (uncommon).
| Copied on clone | Not copied — cloner must do | |---|---| | Schema (DDL) | App user data (unless explicitly seeded) | | RLS policies | Auth provider OAuth secrets | | Functions (handlers + non-secret env) | Function secrets / API keys (cloner provides via preflight) | | App config | **Agents** (record in `agents` table) — bundle as `agents/*.json` in the repo and re-import after clone | | Frontend artifact (with app-id rewritten) | **Storage objects** (only the bucket config replays) | | Repo snapshot (files) | OAuth integration installs |
This makes the **README in the pushed repo** the most important publishing artifact — it has to explain everything that doesn't replay.
Before flipping `visibility: public`, verify all of the following. Each one prevents a class of broken-on-clone bugs.
1. **Repo snapshot exists and is fresh.** Run `butterbase repo push -m "publish v1"` from the project root. Confirm `apps.repo_latest_snapshot` is non-null via `manage_app action: get`. The clone preflight endpoint returns `has_repo: false` if missing, and discovery hides templates without a repo for several sort modes.
2. **README.md at repo root with clone instructions.** No formal "template instructions" field exists — the README in the pushed snapshot **is** the instructions. It MUST cover:
3. **Env-var conventions for auto-mint.** If a function reads `butterbase_api_key`, the clone preflight auto-mints a scoped key for the cloner. Use this convention name exactly so the clone modal can fill it without asking.
4. **No secrets in committed files.** `.butterbaseignore` covers
Claude Code plugin for Butterbase — the AI-Native Backend-as-a-Service. This plugin gives Claude deep knowledge of Butterbase's 42+ MCP tools, guides you through common workflows, and auto-configures the MCP server connection.
Repo: butterbase-ai/butterbase-skills
Use when designing, deploying, or debugging a Butterbase Agent (declarative LLM/tool graph), registering an MCP server for tool use, or wiring access controls…
Use when calling the app's AI gateway from agent tools — chat completions, embeddings, listing models, configuring defaults or BYOK, reading token/cost usage
Use when configuring OAuth providers (Google/GitHub/Apple/X/etc.), setting up post-login auth hooks, tuning JWT lifetimes, or generating service API keys
Use when building a new Butterbase app from scratch, creating a full-stack application, or when the user asks to set up a complete backend with database, auth,…
Use when contributing to the Butterbase codebase, adding new MCP tools, creating API routes, writing migrations, or understanding the monorepo architecture
Use when users report access denied errors, see wrong data, RLS policies are not working, or when troubleshooting Row-Level Security issues in Butterbase