Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Full Stack Agent Workflow: 9 Phases + Next.js Pack

    1

    ```markdown ## One workflow for every change, on any stack AI agents jump straight into code: no plan, no spec, no tests, no record of why.

    $7.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Full Stack Agent Workflow: 9 Phases + Next.js Pack

    Full Stack Agent Workflow: 9 Phases + Next.js Pack

    Example session with this skill installed

    /develop Add a projects page where signed in users can see their projects and create a new one.

    • Read your context and instructions
    • Compiled the full stack agent

    I built the projects page. Signed in users can see their own projects and create new ones. It works end to end in a production build, but it sits on a demo sign in and a JSON file store that I picked myself, and those need your review.

    What I picked for you (non interactive run):

    Spec gate: "Build now, record it as an assumed spec". Three things had never been decided: how users sign in, where projects are stored, and what a project contains. The spec is docs/specs/0001-projects-page/index.md, and it stays Assumed.

    Design: "Extract from existing code". I wrote design.md from the starter page's look (monochrome, Geist font, pill buttons) and added the few colour tokens the page needed in app/globals.css, checked for WCAG AA contrast in light and dark.

    • Git: created branch feat/projects-page and made one commit per milestone.
    • Verify steps: "Save to verify.md", the recommended choice for auth and data work.

    What I assumed:

    Sign in: a demo /sign-in page that asks only for an email (no password). It sets a signed cookie that lasts 7 days and needs a SESSION_SECRET env var. All of it sits behind one file, lib/auth/session.ts, so you can swap in a real auth provider later.

    Storage: a JSON file at .data/projects.json, ignored by git. If two people create a project at the same instant, one save can be lost. That's fine for a demo, not for production.

    Projects: a name (1 to 80 characters) and an optional description (up to 280). The list is newest first, 20 per page, and dates show in UTC.

    Checks: typecheck, lint and next build all pass. I drove the real app in a headless browser, in light and dark mode and at phone width. Every acceptance criterion passed:

    • signed out visitors are redirected to sign in, then sent back
    • each user sees only their own projects
    • the empty state shows
    • creating a project updates the list without a reload
    • bad input shows errors and saves nothing
    • sign out works
    • a forged cookie and a redirect to another site are both rejected

    The first run found one real bug: the session expiry check read the clock in a way Next's caching mode rejects. I fixed it and the rerun was clean. Two other failures in that run were mistakes in my test script, not the app.

    Heads up:

    • You need SESSION_SECRET (32+ characters) in your env. I created a local .env.local with a random one and added .env.example. Without it, the server logs a clear "Invalid environment" error.
    • On a fresh clone, run npx next typegen before npx tsc --noEmit. The existing layout.tsx has the same need.
    • Things I invented for you to replace: the "Demo App" name and its "D" logo, and all the page copy.
    • I installed the small server-only package, which stops server code from leaking into the browser bundle.
    • There's no scope file yet, so I had no feature boxes to tick. The test browser tools are in /tmp/pw-check, outside the repo, and safe to delete.

    Suggested next steps (all optional):

    1. /architect projects page: ratify the assumed sign in, storage, and project model. This is worth doing before you build more on top of these choices.
    2. /check verify projects page, which runs the steps in verify.md.
    3. /test to set up a test runner and lock these flows in, then /check review before a PR.

    You may want to /clear before starting the next feature.

    Save to verify.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    ## One workflow for every change, on any stack

AI agents jump straight into code: no plan, no spec, no tests, no record of why. The next session starts from zero.

**Full Stack Agent Workflow** gives your agent a complete, phase based engineering process. Run only the phases a change needs, in any order. It works with any language or framework: TypeScript, Python, Go, Ruby, PHP, and more.

### The nine phases

| Phase | What it does |
|---|---|
| **/scope** | Turns an idea into a living scope and keeps it current as you ship |
| **/audit** | Writes `AGENTS.md` context files: your stack, conventions, commands |
| **/architect** | Makes a load bearing decision and writes it as a build spec |
| **/develop** | Builds a feature, UI or backend, from its spec |
| **/check** | `verify` runs the real app; `review` reads the code before merge |
| **/test** | Writes a test suite for the code you just changed |
| **/document** | PR text, changelog, release notes, or postmortem from the real diff |
| **/sync** | Keeps `AGENTS.md`, the scope, and spec statuses current |
| **/debug** | Finds the root cause of a bug, fixes it, and adds a regression test |

### State lives in files, not chat

Scopes, specs, `AGENTS.md`, and tests are saved in your repo, so work picks up across sessions and your whole team shares the same workflow.

### Bonus: Next.js pack

On a Next.js project, every phase automatically loads extra Next.js rules:

- **Server first:** `"use client"` only on the smallest interactive leaf
- **Secure mutations:** every Server Action validates input and checks auth itself
- **Correct caching:** works with Cache Components (`"use cache"`, `cacheTag`) or classic segment config, and always revalidates what changed
- **Version aware:** follows your installed `next` version's conventions
- **Production checks:** runs `next build` and `next start`, checks hydration warnings and loading, error, and not found states
- **Bug table:** hydration mismatches, stale data, missing `NEXT_PUBLIC_`, swallowed `redirect()`, and more

On any other stack, the pack stays out of the way.

    Requires local file storage for specs and AGENTS.md. Verification phases require a working local environment to run the app. Next.js pack applies only when next is in package.json.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed today

    Requires

    Git
    Your project's own runtime and test tools (Node.js, Python, etc.)
    Playwright (optional, for browser checks)
    Python
    Node.js
    macOS

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Browser
    Network Access

    File Scopes

    AGENTS.md
    docs/**

    Writes app code, tests, and workflow files (docs/scope, docs/specs, AGENTS.md) inside your project. Never reads or writes secret values. No external accounts needed.

    Listedtoday

    What's inside

    Frequently Asked Questions

    It gives your AI agent a nine phase engineering process (scope, audit, architect, develop, check, test, document, sync, debug), so changes go from idea to shipped, verified, documented code.

    Any. It reads your project and follows its conventions. Next.js projects also get the built in Next.js pack.

    Any agent that supports the SKILL.md standard: Claude Code, Codex CLI, Cursor, Gemini CLI, VS Code Copilot, and more.

    Ask your agent for a phase, like `/audit` on an existing project or `/scope` for a new idea. Run only the phases a change needs. For a bug, go straight to `/debug`.

    Extra rules that load only on Next.js projects: Server and Client Component boundaries, secure Server Actions, caching, production build checks, Next.js testing, and a common bugs table.