Skip to content
Agentmart
Submit a tool

product-build

End-to-end product build conductor.

Skill by TheNonFungibleHuman

IndexedMIT
Write a review

Chalk Labs

Need help deploying product-build?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/thenonfungiblehuman-product-build

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

End-to-end product build conductor. Use when the user asks to build, ship or scaffold an app, product, MVP, feature, or hackathon project; says "build me", "make an app", "ship this", or wants a design-to-code build. Runs grill -> spec -> design gate -> plan on pro -> implement on flash -> verify -> review -> handoff, with the Impeccable detector as a hard anti-slop gate. Do not use for one-line fixes or trivial edits.

Use cases: design

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Nov 1, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/product-build

What's inside

Skill name
product-build
Description
End-to-end product build conductor. Use when the user asks to build, ship or scaffold an app, product, MVP, feature, or hackathon project; says "build me", "make an app", "ship this", or wants a design-to-code build. Runs grill -> spec -> design gate -> plan on pro -> implement on flash -> verify -> review -> handoff, with the Impeccable detector as a hard anti-slop gate. Do not use for one-line fixes or trivial edits.
License
MIT
Instructions
164 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Product build (conductor)

You own the pipeline. Do not wait to be told to delegate, to use a skill, or to
route a model; decide and proceed. Ask the user only product questions, never
permission for each step.

## 0. Classify

If the request is a new surface, feature, app, or MVP: run this pipeline. If it
is a one-line fix or a question, skip it and use the normal workflow.

## 1. Grill (one round, max 8 questions)

Ask only what changes the build:

1. Who is it for, and what do they do with it?
2. What are the 3 must-haves, in order?
3. What is explicitly out of scope?
4. Stack/runtime/deploy constraints?
5. Existing brand, assets, references, screenshots?
6. Definition of done (the exact demo behaviour + tests)?
7. Data/auth/payments/integrations?
8. Deadline or budget constraint?

If the user says "just build", proceed with an **Assumptions** list and say so.
Do not ask a second round of questions.

## 2. Spec and docs

Write at the repo root:

- `SPEC.md` — goal, users, must-haves, non-goals, acceptance checks.
- `DESIGN.md` — after the design gate (step 3).
- `AGENTS.md` — repo conventions and run/test commands.
- ADRs only for irreversible decisions.

Use the library when useful: `to-spec`, `to-tickets`, `domain-modeling`,
`grill-with-docs`. Search with the `skill-router`; load at most 3 skills.

## 3. Design gate (no UI code before this)

**Ask for references first.** Ask: “Do you have references — URLs, screenshots,
notes — or should I pull from Inspo?” Pass any with `-References`; if none, the
lane uses Inspo (free, MIT). Use `-NoInspo` only if they want no outside
inspiration. Steal like an artist: synthesize 2–3 sources; never copy a layout
1:1, copy, logos, or brand assets.

1. Produce **2–3 genuinely different directions** on the **pen.dev canvas**
   through the design lane:
   ```powershell
   powershell -File "$env:USERPROFILE\.grok\skills\skill-router\scripts\design-lane.ps1" `
     -Brief "<direction brief>" -Cwd . -Out .\design\directions\<name>
   ```
   No HTML/CSS mockups and no review server. The lane exports `preview.png`.
2. **AI review first:** check each direction against the canvas checklist in
   `rules/02-design-gate.md` (hierarchy, type, palette, contrast, spacing, one
   CTA, banned tells). Fix or reject in the canvas before showing the user.
3. The user picks in Pencil (or you pick only if they delegated).
4. Write `DESIGN.md`: palette, type scale, spacing, components, motion, chosen
   frame.
5. **Choose the component library after the pick**, from the stack and the job:
   React/Next → shadcn/ui + Radix + Tailwind; Vue → Nuxt UI; Svelte → Skeleton;
   Angular → Spartan/NG-ZORRO; marketing-only → none unless reused. One library,
   no component MCP, no full kit dump. Record the choice and why in `AGENTS.md`
   or `DESIGN.md`.
6. Then — and only then — implement.

Surface → skill (via `skill-router`):

- Product UI / app shell / dashboard → **Impeccable** (`shape` → `craft`) +
  Jakub `better-ui` / `better-typography` / `better-layout`.
- Component states → Jakub `break` / `state-machine`.
- Motion → `emil-design-eng`.
- Landing / marketing / portfolio → **Taste v2** (after direction approval).
- Copy → Jakub `better-writing`; color system → `better-colors`.
- Figma/design image → code → Jakub `build-design`.

One primary + one gate. No rival reviewers. No pack enabling.

## 4. Plan on pro

Spawn the `plan` subagent (config pins it to `deepseek-pro`) with `SPEC.md` and
`DESIGN.md`. Require:

- Goals / non-goals
- Ordered steps with files/areas and acceptance checks
- Risks and open questions
- The smallest vertical slice to build first

Then implement that slice on Flash. Do not plan the whole product on Flash.

## 5. Implement on Flash

- Main session or a `general-purpose` child on `deepseek-flash`; keep bulk work
  off the pro seat.
- Load 1–3 skills via `skill-router`; delegate heavy skill bodies to a child and
  take back a short summary.
- Follow `DESIGN.md`; write tests with the slice; run them.
- End each change with the evidence line: files changed + exact test command +
  result (or the blocker).

## 6. Verify (hard)

- `rules/01-verification.md` + `tdd`: symbol sweep, project env first,
  neighborhood sweep, evidence line.
- The design detector (Stop hook) blocks completion while changed UI files have
  findings. Fix them or add a justified waiver. Do not disable the gate.
- For the repo itself, install the pre-commit gate once:
  ```powershell
  powershell -File "$env:USERPROFILE\.grok\skills\skill-router\scripts\install-design-guard.ps1" -Repo .
  ```
  It scans staged UI files and blocks the commit on findings.
- Visual proof: screenshot desktop + mobile and compare with the chosen
  direction. No screenshot, no "done".

## 7. Review on pro

Spawn **one** reviewer (`general-purpose`, `model: deepseek-pro`) with
`code-review` or `diagnosing-bugs`. Give it the diff, the spec, and the design

From TheNonFungibleHuman/multi-model-agent-harness, MIT.

More from TheNonFungibleHuman/multi-model-agent-harness

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Oct 19, 2026 (13 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from TheNonFungibleHuman

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get product-build running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying product-build?

CallGet help
Cite this page

product-build skill. Agent Mart. https://agentmart.dev/skills/thenonfungiblehuman-product-build (last updated 2026-10-24)