agentleFS
TomSzenessyTomSzenessy/AI_TEMPLATE/.agents/skills/product-kickoff/SKILL.md

product-kickoff

Skill0 starsChanged 11 days agoMITView on GitHub

Turns 'build me an app/site/game' into confirmed decisions before any code: asks the owner a short round of questions with recommendations (users, platforms, stack, visual style, onboarding, data), shows two or three mockup directions, and records the choices. Use at the start of every new product or major reshaping.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read VISION.md and docs/STACK-DECISION.md.

    line 16
  2. Distinctive:

    Ask everything in one message (use the host's question tool when it has one).

    line 42
  3. Distinctive:

    Ask a second round only if an answer opens a new high-impact choice.

    line 72
  4. Distinctive:

    Only an explicit "use your picks" counts as approval.

    line 77
  5. Very specific:

    Feature list: write docs/product/features.csv (feature, area, priority, status, evidence, acceptance, source) from the owner's goal, the …

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Product kickoff

Outcome: before any product code, the owner has confirmed who it is for, the platforms, the stack, and a visual direction they have seen; research, a feature list with acceptance criteria, and the decisions are recorded where every later agent finds them. Then make next drives the build to launch.

1. Understand before asking

  • Read VISION.md and docs/STACK-DECISION.md. If they are already accepted, this is a reshape: ask only about what changes, then update both and record the change as an ADR; never silently overwrite a confirmed decision.
  • Restate the idea in one sentence and name the riskiest assumption.
  • Size it: prototype (days, throwaway or demo), product (an MVP for real users), or platform (many users, teams, or services). The tier sets how much stack-foundation and docs/engineering.md "Built to scale" apply; never build platform machinery for a prototype.

2. Research the space (always)

Delegate to the researcher role (web search and browser) and write docs/product/research.md with cited URLs (make check needs at least three) and an <!-- index: extend | ... | ... --> line under its title:

  • Comparable products: the three to five best, what each does well, and what to borrow. For a named reference, run product-recon.
  • What users hate: review-mining on their reviews; each complaint is a feature or a differentiator.
  • Design references: apps or sites whose look and feel fit the brief (screenshots in ignored .agent/references/), and what to borrow from each.
  • Our angle: the one thing this product will do better.

Bring the findings into the questions as recommendations.

3. Ask one round of questions

Ask everything in one message (use the host's question tool when it has one). Give each question options and your recommendation, so "go with your picks" is a valid answer. Ask only what changes direction:

  1. Users and goal: who, which problem, what success looks like (one metric).
  2. Platforms: web, iOS, Android, desktop; offline needed?
  3. Stack: an existing codebase or a stack the owner already uses wins. Otherwise recommend from platforms, team skills, tier, and constraints. Defaults to start from (before scaffolding, have the researcher role or Context7 confirm the choice is still recommended and get current versions):

    NeedDefaultConsider instead
    Web appReact + Next.js + Tailwind (+ shadcn/ui)SvelteKit; Astro for content sites
    iOS + Android, one codebaseExpo (React Native)Flutter for highly custom, pixel-identical UI
    Best native feel, one platformSwiftUI / Jetpack Compose—
    Desktop appTauri (small, web UI)Electron for Node-heavy needs; native toolkits
    GameWeb canvas or Phaser for small 2D; Godot (open source) for 2D/3DUnity or Unreal for console or high-end 3D (check license terms)
    API, CLI, or data/MLThe team's strongest language; Python for data and ML; TypeScript or Go for services—
    BackendManaged (e.g., Supabase or Firebase) for prototype and productOwn services with Postgres, queues, and caches at platform tier
  4. Visual style: platform-native (Apple HIG, Material 3) or a custom brand; two or three apps or sites they like and why; mood words; light, dark, or both; existing logo or colours.
  5. First-run and core flow: what a new user does in the first minute; is sign-up needed before value?
  6. Data, accounts, payments, privacy: sign-in method, payments, children or health data, regions; personal data means docs/privacy.md.
  7. Constraints: budget, deadline, hosting, must-use services, languages.

Ask a second round only if an answer opens a new high-impact choice. Never assume answers: if you cannot ask (headless run, tool unavailable) or the owner has not answered, write the questions with your recommendations in your reply (and the kickoff issue when one exists), then stop and wait for answers. Once the owner has answered, never stop again before make next reaches launch. Only an explicit "use your picks" counts as approval.

4. Show mockups before building

  • Produce two or three clearly different directions for the one or two most important screens (usually first-run and the main screen), with the same content so only the design differs. Static HTML + Tailwind renders anywhere; use platform previews when the stack is native.
  • Render them at phone and desktop sizes with the browser tool (Playwright MCP), check them against ux-quality yourself, then show the screenshots to the owner with a one-line rationale per direction. If make capabilities shows no browser, give the owner the file paths to open instead.
  • Iterate on the chosen direction once or twice. Keep mockups in ignored .agent/mockups/ unless the owner wants them in the repository.

5. Define the product and plan delivery

  • Feature list: write docs/product/features.csv (feature, area, priority, status, evidence, acceptance, source) from the owner's goal, the research, and the competitor complaints, then append the rows of production-features.csv (this folder) that apply. It is the one matrix: product-recon and review-mining append rows here too (source says who). "Complete" means every must row is yes with evidence; a demo-sized list is a defect.
  • Architecture sketch: data model, main API or screen contracts, auth, hosting, and how it grows (the "Built to scale" section of docs/engineering.md). Record hard-to-reverse choices as ADRs.
  • Walking skeleton first: the thinnest end-to-end path (one screen, one request, one stored record), deployed early.
  • Vertical slices: split the MVP into user-visible slices, each shippable and verifiable on its own; one issue per slice under an epic issue, ordered by risk and value. No Markdown roadmap.
  • Success metric: how the metric from question 1 will be measured after launch, with privacy-respecting analytics only where it is needed.

6. Record and hand off

  • VISION.md (status accepted), docs/STACK-DECISION.md (decision, rejected alternatives, and why).
  • make new KIND=doc NAME=design GROUP=design DESC="Visual direction, tokens, and UX principles; before any UI work", then fill it with the chosen direction, references, and tokens (the design-tokens conventions and contrast checker apply).
  • File the epic and slice issues, then run the stack-foundation skill for the walking skeleton; its surface's quality oracle includes the ux-quality review. Then follow make next slice by slice until it reports the launch phase; never stop at a green make done while must features are open.

Done when

  • The owner explicitly confirmed platforms, stack, and one mockup direction.
  • docs/product/research.md cites the comparable products, complaints, and design references; docs/product/features.csv covers them plus the production rows, each must row with an acceptance criterion.
  • An epic with ordered slice issues exists, starting with the walking skeleton.
  • VISION.md, docs/STACK-DECISION.md, and docs/design.md record them and make check passes.

Never

Scaffold a framework before confirmation, ask questions one at a time across many turns, or copy a reference product's names, assets, or copy.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.