Superstack Tutorial — Step-by-Step Guide

0

Prerequisites

2 min

Goal

Make sure you have everything needed before installing Superstack.

RequiredNode.js & npm

Check version

$ node --version # need v18+ v20.11.0 ✓

RequiredAn AI agent runtime

You need at least one of these installed and working:

$ claude --version # Claude Code (recommended) # OR $ codex --version # OpenAI Codex

[+]

Recommended: Claude CodeThis tutorial uses Claude Code throughout. Install it with npm install -g @anthropic-ai/claude-code.

OptionalSolana CLI (for build/deploy steps)

$ sh -c "$(curl -sSfL https://release.solana.com/stable/install)" $ solana --version

1

Install Superstack

1 min

Goal

Install all Superstack skills into your agent's context directory with a single command.

StepRun the installer

Terminal — run from any directory

$ bash <(curl -sSf https://solana.new/install)

Expected output

Detecting agent runtimes...
✓ Claude Code found at ~/.claude/
✓ Codex found at ~/.codex/
Installing 24 skills...
✓ validate-idea
✓ brand-design
✓ create-pitch-deck
... (24 total)
✓ Superstack installed successfully

VerifyConfirm installation

$ ls ~/.claude/skills/ validate-idea.md brand-design.md create-pitch-deck.md ...

[+]

No restart neededSkills are available immediately in any new Claude Code or Codex session. Open a new terminal and they're ready.

2

Navigate the Skill Library

3 min

Goal

Get a full overview of every available skill, organised by phase.

StepOpen Claude and ask for the skill list

In Claude Code (or any agent)

Claude: tell me all skills available in Superstack
and help me navigate across skills

Expected output

Loading /navigate-skills...

IDEA PHASE
• /find-next-crypto-idea
• /validate-idea
• /competitive-landscape
• /defillama-research
• /colosseum-copilot

BUILD PHASE
• /scaffold-project
• /build-defi-protocol
• /frontend-design-guidelines
• /brand-design
• /build-mobile

LAUNCH PHASE
• /create-pitch-deck
• /marketing-video
• /video-craft
• /deploy-to-mainnet

DESIGN & META
• /design-taste
• /roast-my-product
...

[+]

Pro tipStart every session with this prompt. The list is your map — pick the skill that matches where you are in your project lifecycle.

3

Find a Product Idea

5 min

Goal

Either use Superstack to generate ideas, or bring your own idea from solana.new's top ideas list.

Generate an idea

Browse top ideas

StepUse /find-next-crypto-idea

Claude: /find-next-crypto-idea # or with constraints: Claude: /find-next-crypto-idea — I want something in DeFi,
medium complexity, that I can demo at a hackathon

The skill will generate 5–10 ranked ideas with market rationale, existing alternatives, and an initial feasibility estimate.

AlternativeBrowse solana.new for curated ideas

Visit solana.new and copy any idea you find interesting. The site maintains a curated list of the most-wanted builds in the Solana ecosystem.

For this tutorial we'll use Power Perps:

# Paste the idea into your Claude session: Claude: Build a perpetuals trading protocol with power-law
pricing mechanics targeting retail traders on Solana.

4

Validate the Idea

8 min

Goal

Get an objective, scored assessment of your idea before committing to build it.

/validate-idea

~2 min to runOutputs HTML report6 scoring dimensions

Evaluates your idea on six dimensions and produces a detailed HTML report with scores, risks, and recommendations. The report includes an "integration-first" section that identifies existing Solana protocols you could build on top of instead of building from scratch.

Produces: HTML reportScore cardRisk summaryIntegration map

StepRun the validation

Claude: /validate-idea # paste your problem statement when prompted: Build a perpetuals trading protocol with power-law pricing mechanics targeting retail traders on Solana.

Output (excerpt)

Validation complete. Opening report...

SCORECARD
Demand Signal ████████░░ 82% Strong ✓
Competitive Risk █████░░░░░ 55% Medium
Build Feasibility ███████░░░ 72% High ✓
Crypto Necessity █████████░ 90% Essential ✓
Target User ██████░░░░ 60% Needs work
Integration-First ████████░░ 80% Yes ✓

KEY RECOMMENDATION
Build on top of Drift Protocol or Jupiter Perpetuals
rather than building a new AMM from scratch.

Demand Signal

82%

Competitive Risk

Med

Build Feasibility

72%

Crypto Necessity

90%

Target User Clarity

60%

Integration-First

80%

What to do with your score

Overall score ≥ 75% → Proceed to build phase

Score 50–74%, addressable issues → Fix top issues, re-run /validate-idea

Score <50% or "no crypto necessity" → Pivot; run /find-next-crypto-idea

Low Integration-First score → Research existing protocols first

5

Colosseum Hackathon Alignment

3 min

Goal

Map your idea to the Colosseum judging rubric and choose the best prize track to target.

StepRun /colosseum-copilot

Claude: /colosseum-copilot # then describe your idea when prompted

Output (excerpt)

COLOSSEUM ALIGNMENT

Recommended track: DeFi Innovation
Win probability: Medium-High

Rubric fit:
Innovation ████████░░ 78%
Technical Depth ████████░░ 80%
Market Potential █████████░ 88%
Presentation Ready ██████░░░░ 65% ← needs work

Key differentiator to emphasise: power-law payoff
structure not seen in existing Solana perps.

[!]

Presentation scoreA low Presentation score here means your pitch deck needs work — not your product. Use /create-pitch-deck in the launch phase to address this.

6

Brand & Design System

5 min

Goal

Choose a visual identity and generate CSS design tokens that the UI generation step will use.

StepTrigger brand design (via frontend skill)

Claude: /frontend-design-guidelines — build the front end
for this perps trading idea using mock data to start

The skill automatically invokes /brand-design first. It will generate 6 colour palette options as an HTML file and ask you to choose one.

StepChoose your palette

Open the generated HTML file in your browser. Six palettes are shown. For a trading protocol, consider:

Palette	Best for	Feeling
`whale-blue`	Trading / DeFi	Trustworthy, institutional
`solana-gradient`	Native Solana products	Modern, energetic
`midnight-gold`	Premium / high-value	Exclusive, polished
`void-green`	Dev tools / data-heavy	Technical, focused

Claude: let's go with whale blue

[+]

The palette generates CSS variablesAfter you choose, the brand skill outputs :root CSS variables for colour, typography scale, and spacing. These are used automatically by the frontend generation step.

7

Build the Front End

8 min

Goal

Generate a complete, styled trading interface with mock data — ready for visual review before any contracts are written.

[x]

Always use mock data firstTell the skill explicitly to use mock data. This lets you review and approve the UI before it drives your Anchor account structure. Changing account layouts after writing contracts is expensive.

StepGenerate the UI

Claude: now build the full front end for the Power Perps
trading interface — use mock data throughout,
whale blue theme, run on localhost:3000

What gets generated

✓ Template 1: Dashboard view (portfolio + open positions)
✓ Template 2: Trading panel (order book + P&L chart)
✓ Template 3: Position management modal
✓ CSS variables from whale blue brand
✓ Mock data layer (positions, trades, chart data)
✓ Responsive layout

Running on http://localhost:3000 ✓

ReviewVisually approve the UI

Open http://localhost:3000 and review each template. Ask yourself:

Does the layout reflect how traders actually work?
Is the data hierarchy correct (what's prominent vs secondary)?
Are the right interaction patterns present (modals, confirmations)?
Does the brand feel match the target user?

Request specific changes before moving on. Layout changes here cost nothing. Account layout changes after contracts cost days.

NoteWhat the skill also infers

Based on the UI it generated, the skill automatically infers and lists:

Anchor account structs — field names and types
Instruction signatures — open_position, close_position, etc.
PDAs needed — user position account, global state, etc.

This list feeds directly into the scaffold step.

8

Scaffold the Smart Contract

5 min

Goal

Generate the Anchor program skeleton that matches the UI's data shape, so no account layout redesigns are needed.

StepFor simple projects — /scaffold-project

Claude: /scaffold-project

Generates full project directory structure, package.json, Anchor.toml, and basic program skeleton.

StepFor DeFi protocols — /build-defi-protocol

Claude: /build-defi-protocol — based on the front end we built,
scaffold the Anchor program for Power Perps with
the account structs you inferred from the UI

What gets scaffolded

programs/power-perps/src/
├── lib.rs ← program entry point
├── state.rs ← account structs (Position, GlobalState, ...)
├── instructions/
│ ├── open_position.rs
│ ├── close_position.rs
│ └── update_price.rs
├── errors.rs ← custom error codes
└── constants.rs ← program constants

tests/power-perps.ts ← Anchor test suite

[+]

From hereThe scaffold gives you the structure. Fill in the business logic (the power-law pricing math) — that's your actual innovation, and no AI skill replaces it.

9

Create the Pitch Deck

5 min

Goal

Generate a complete, hackathon-ready HTML pitch deck with rubric-projected scores — with minimal editing needed after.

/create-pitch-deck

~90 secondsHTML output8 slidesRubric-scored

Generates 8 slides mapped to the Colosseum judging rubric. Asks 3 questions before generating. Includes projected scores per rubric dimension and flags 2 placeholders you need to swap before submitting.

Produces: 8-slide HTML deckRubric projectionPlaceholder list

StepGenerate the pitch deck

Claude: /create-pitch-deck

Answer the 3 questions when prompted:

What is this deck for? → Colosseum What tone of writing on the slides? → Technical but accessible Team size and stage? → 2-person team, prototype stage

After generationSwap the 2 placeholders

The skill will tell you exactly what to replace. Typically:

[TEAM_PHOTO] → actual team photo or headshots
[DEMO_URL] → live demo link or Vercel preview URL

[!]

Don't skip the placeholdersJudges notice blank or obviously placeholder content. The deck is 95% ready — the 2 swaps are the last 5%.

10

Roast Your Product

3 min

Goal

Get a brutally honest pre-submission quality check — find the issues before the judges do.

StepRun /roast-my-product

Claude: /roast-my-product

Example roast output

ROAST REPORT

"This dead protocol is proof of demand with a
formidable name and zero engine code."

SCORE: 45 / 110

ISSUES (ranked by severity):
1. [CRITICAL] Common sense detected in target user section
→ "retail traders" is too broad. Define: Drift users
with $10k+ portfolios wanting leveraged exotic plays.
2. [HIGH] UX red flags on position entry flow
→ 4-step confirmation is 3 steps too many
3. [MEDIUM] Token model is unclear
→ Is there a token? Why? Answer clearly or remove claim.

VERDICT: Ship the prototype, get 5 real users,
validate demand before writing more contract code.

How to interpret the roast score

80–110 — Green → Submit as-is. You're in good shape.

60–79 — Yellow → Fix the top 2 issues, re-run roast.

40–59 — Orange → Address all critical issues before submitting.

<40 — Red → Pivot the positioning. The idea may be fine but the story isn't.

[+]

A low score is not failureIt's a prioritised fix list. A roast score of 45 with a clear fix list is more actionable than a 70 with vague feedback.

11

Marketing Video

3 min

Goal

Generate a script and storyboard for a 60–90 second explainer video to share on X/Twitter.

StepGenerate video script

Claude: /marketing-video — Power Perps, targeting Solana DeFi
traders, 60 seconds, share on X

The skill outputs a shot-by-shot script with narration, on-screen text, and visual cues. Use it as a brief for a screen recording or a designed video.

[i]

/video-craftFor a more detailed production brief — including hook structure, pacing, and platform-specific optimisations — use /video-craft after /marketing-video.

12

Deploy to Mainnet

10 min (when ready)

Goal

Follow the Superstack deployment checklist to ship safely — program verification, upgrade authority, and monitoring.

StepRun the deployment checklist

Claude: /deploy-to-mainnet

The skill walks you through:

Program verified on devnet with full test suite passing
Anchor.toml updated with mainnet cluster
Upgrade authority transferred to multisig
Buffer account funded for deployment
Program deployed and verified on-chain
Frontend pointing to mainnet program ID
Post-deploy smoke test passed
Monitoring set up (e.g. Helius webhooks)

[x]

Upgrade authority is criticalBefore mainnet, transfer the upgrade authority to a multisig. A single keypair as upgrade authority is a single point of failure and a red flag for serious users.

→

All Skills Reference

Skill	Phase	What it produces
`/navigate-skills`	Meta	Full skill catalogue, organised by phase
`/find-next-crypto-idea`	Idea	Ranked idea list with market rationale
`/validate-idea`	Idea	Scored HTML report across 6 dimensions
`/competitive-landscape`	Idea	Competitor matrix + opportunity map
`/defillama-research`	Idea	Live DeFi market data summary
`/colosseum-copilot`	Idea	Rubric alignment + prize track recommendation
`/brand-design`	Build	6 palette options → CSS design tokens
`/frontend-design-guidelines`	Build	Full HTML UI (3 templates) + inferred account layout
`/scaffold-project`	Build	Project directory + Anchor skeleton
`/build-defi-protocol`	Build	Anchor program + account structs + tests
`/build-mobile`	Build	React Native + Mobile Wallet Adapter scaffold
`/solana-beginner`	Build	Commented scaffold for newcomers
`/design-taste`	Design	UI critique with specific improvements
`/roast-my-product`	Launch	Scored critique + prioritised fix list
`/create-pitch-deck`	Launch	8-slide HTML deck + rubric projection
`/marketing-video`	Launch	Script + storyboard for 60–90s video
`/video-craft`	Launch	Platform-optimised video production brief
`/deploy-to-mainnet`	Launch	Deployment checklist + verification steps

→

Contributing to Superstack

Superstack is completely open source. If you built something useful with it, consider contributing the skill back.

How to contribute4 steps to add a skill

$ git clone https://github.com/solana-labs/superstack $ cd superstack/skills # Create your skill file following the template $ cp _template.md my-new-skill.md # Test locally $ cp my-new-skill.md ~/.claude/skills/ # Open a PR with demo prompt + expected output

[+]

What makes a good skillSolana-specific, produces a tangible artifact (HTML/code/report), guides the AI through a multi-step workflow, and includes expert context the AI wouldn't have by default.

[i]

Issues and PRs welcomeUse the skill-request label to request a new skill. Use bug if a skill's output is wrong or outdated. The community reviews and merges quickly.

Build a Solana Product
Idea to Launch

Prerequisites

Install Superstack

Navigate the Skill Library

Phase 1: Idea

Find a Product Idea

Validate the Idea

What to do with your score

Colosseum Hackathon Alignment

Phase 2: Build

Brand & Design System

Build the Front End

Scaffold the Smart Contract

Phase 3: Launch

Create the Pitch Deck

Roast Your Product

How to interpret the roast score

Marketing Video

Deploy to Mainnet

All Skills Reference

Contributing to Superstack

Build a Solana ProductIdea to Launch

Prerequisites

Install Superstack

Navigate the Skill Library

Phase 1: Idea

Find a Product Idea

Validate the Idea

What to do with your score

Colosseum Hackathon Alignment

Phase 2: Build

Brand & Design System

Build the Front End

Scaffold the Smart Contract

Phase 3: Launch

Create the Pitch Deck

Roast Your Product

How to interpret the roast score

Marketing Video

Deploy to Mainnet

All Skills Reference

Contributing to Superstack

Build a Solana Product
Idea to Launch