building-edgespark-apps — Claude Code skill: what it does, install
Guides coding agents working on EdgeSpark applications. Ensures correct use of the CLI, server SDK types, database batching, and browser authentication.
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a claude-codeWhat building-edgespark-apps does
This skill sets development constraints for repositories containing edgespark.toml. It defines the required file inspection sequence, forbids manual changes inside src/__generated__/, and requires running edgespark pull types to refresh placeholder interfaces.
The prompt enforces concrete patterns across the stack. Client code must rely on @edgespark/web, authUI.mount() for UI login flows, and es.api.fetch() for API communication. On the backend, it replaces db.transaction() with db.batch() and requires schema updates through migration files rather than ad-hoc DDL.
Executing multiple edgespark commands concurrently breaks the workflow, so the skill strictly blocks parallel CLI operations.
What it helps with
- configure client authentication with
@edgespark/webandauthUI.mount() - refresh SDK interfaces via
edgespark pull types - group database writes using
db.batch() - generate presigned PUT URLs for client-side storage uploads
Example prompts for the agent
- Mount the login UI inside
#authusing@edgespark/weband redirect to/dashboard - Create a new database migration and sync types with
edgespark pull types - Add a Hono endpoint that checks
auth.userimported fromedgespark/http
How to install building-edgespark-apps
1Skills CLI (Vercel)
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a claude-code2Manually: put the files in place
~/.claude/skills/building-edgespark-apps/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/building-edgespark-apps/
1Skills CLI (Vercel)
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a codex2Manually: put the files in place
.agents/skills/building-edgespark-apps/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a cursor2Manually: put the files in place
.cursor/skills/building-edgespark-apps/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install edgesparkhq/agent-skills building-edgespark-apps2Manually: put the files in place
.github/skills/building-edgespark-apps/SKILL.md1Skills CLI (Vercel)
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a gemini-cli2Platform CLI
gemini skills install https://github.com/edgesparkhq/agent-skills --path skills/building-edgespark-appsManually: put the files in place
.devin/skills/building-edgespark-apps/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about building-edgespark-apps
How to install building-edgespark-apps in Claude Code?
Run npx skills add edgesparkhq/agent-skills --skill building-edgespark-apps -a claude-code. The file will be placed at ~/.claude/skills/building-edgespark-apps/SKILL.md.
What does building-edgespark-apps do?
It provides implementation rules for building EdgeSpark applications. It guides the model through CLI flags, auth routes, schema migrations, and client SDK conventions.
How does the skill handle file uploads?
It prevents streaming files through the Worker runtime and directs the model to create presigned S3 PUT URLs for direct client uploads.
Source text
SKILL.md · MIT
---
name: building-edgespark-apps
description: Build and modify EdgeSpark apps. Use when a project has edgespark.toml, the user mentions EdgeSpark, or work involves the edgespark CLI, server SDK types, storage/auth/database workflows, deployment, or @edgespark/web.
metadata:
author: EdgeSpark
version: "0.0.4"
---
# EdgeSpark App Development
Use this skill for EdgeSpark-specific implementation and workflow decisions.
This skill is not EdgeSpark documentation. For exact contracts, read source, generated types, CLI help, and docs/Mintlify MCP. Use this skill for workflow, guardrails, and bug-prevention.
The reliable public surface in this repo is:
- the `edgespark` CLI
- scaffolded project structure from `edgespark init`
- generated `src/__generated__/edgespark.d.ts`
- generated `src/__generated__/server-types.d.ts`
- the `@edgespark/web` browser SDK
Use `@edgespark/web` and `authUI.mount()` as the default browser auth path for this repo unless custom forms are explicitly requested.
## Read Order
Read only what is needed for the task:
1. `edgespark.toml`
2. repo or project agent instruction file (`AGENTS.md`, `CLAUDE.md`, or `GEMINI.md`)
3. `src/__generated__/edgespark.d.ts`
4. `src/__generated__/server-types.d.ts`
5. `src/defs/index.ts`, `src/defs/db_schema.ts`, `src/defs/db_relations.ts`, `src/defs/runtime.ts`, `src/defs/storage_schema.ts`
6. `node_modules/@edgespark/web/dist/index.d.ts` when installed for exact browser SDK types
7. `node_modules/@edgespark/web/README.md` when installed for managed auth appearance variable meanings and defaults
Then load the specific reference you need:
- Day-to-day development workflows by surface: [dev-workflow.md](references/dev-workflow.md)
- Scaffold layout and generated-file rules: [project-structure.md](references/project-structure.md)
- Error-prone server-side usage patterns: [server-patterns.md](references/server-patterns.md)
- Small web usage patterns for `@edgespark/web`: [web-patterns.md](references/web-patterns.md)
- Auth config, OAuth providers, callback URLs, managed auth theming, and Google One Tap: [auth-patterns.md](references/auth-patterns.md)
## Hard Rules
- Run `edgespark <command> --help` before assuming flags or exact behavior.
- Run `edgespark` commands on behalf of the user. Only hand off steps that explicitly require a human browser action.
- Never run multiple `edgespark` commands in parallel.
- Treat scaffolded `src/__generated__/edgespark.d.ts` and `src/__generated__/server-types.d.ts` as placeholders until `edgespark pull types` populates them.
- Do not edit files under `src/__generated__/`.
- Use `@edgespark/web` for new browser code.
- Use `es.api.fetch()` for app API calls, not bare `fetch()` to same-origin app routes.
- Use `authUI.mount()` for managed auth UI unless custom forms are explicitly requested.
- For managed auth theming, use `appearance.theme` and `appearance.variables` from `@edgespark/web`; do not tell users to edit SDK CSS for routine light/dark or brand theming.
- For custom browser auth flows, use `client.auth` from `@edgespark/web`, not manual `/api/_es/auth/*` calls.
- Import `auth` from `edgespark/http`, not `edgespark`.
- Auth is a managed service at `/api/_es/auth/`. OAuth callback URLs use `/api/_es/auth/callback/<provider>`, not `/api/auth/`.
- Treat `/api/_es/auth/*`, storage provider details, and deployment internals as platform implementation details unless the user is explicitly debugging them.
- Do not import runtime SDK values from `edgespark` inside `src/defs/**`.
- Use `db.batch()` instead of `db.transaction()`.
- Use migration workflow for schema changes. Do not use DDL through `edgespark db sql`.
- Store S3 URIs in the database and return presigned URLs to clients.
- For client-originated uploads, generate presigned PUT URLs instead of streaming files through the Worker.
- Update `src/defs/runtime.ts` before using `vars.get()` or `secret.get()`.
- Use `edgespark ... --help` for exact command syntax instead of duplicating help text in this skill.
- If exact behavior is unclear, prefer source code, generated types, or docs MCP over guessing.
## Default Workflow
For the operational workflow by area, read [dev-workflow.md](references/dev-workflow.md).
### Existing project
1. Read generated type files first.
2. Read the relevant defs files before changing schema, storage, or runtime keys.
3. Read the web SDK types before touching auth or browser API code.
### Fresh scaffold
1. Inspect `edgespark.toml` to confirm server-only vs full-stack layout.
2. If generated files are placeholders, run `edgespark pull types` before making SDK assumptions.
3. Follow the scaffolded root, `server/`, and `web/` agent instruction files for package boundaries.
## When Stuck
1. Read the generated type files again before assuming an API shape.
2. Run the relevant `edgespark ... --help` command before guessing flags.
3. Use docs/Mintlify MCP for product documentation details.
## Quick Start
Server:
```ts
import { db, storage, vars, secret, ctx } from "edgespark";
import { auth } from "edgespark/http";
import { posts, buckets } from "@defs";
import { Hono } from "hono";
import { eq } from "drizzle-orm";
const app = new Hono()
.get("/api/posts", async (c) => {
return c.json(await db.select().from(posts));
})
.post("/api/posts", async (c) => {
const data = await c.req.json();
const [post] = await db.insert(posts)
.values({ ...data, user_id: auth.user!.id })
.returning();
return c.json(post, 201);
});
export default app;
```
Web:
```ts
import { createEdgeSpark } from "@edgespark/web";
import "@edgespark/web/styles.css";
const es = createEdgeSpark();
es.authUI.mount(document.getElementById("auth")!, {
redirectTo: "/dashboard",
});
const res = await es.api.fetch("/api/posts");
const posts = await res.json();
```
© EdgeSpark · MIT · edgesparkhq/agent-skills