Superteam Brasil · Official Skill Startup Builder

positioning

Write the one-liner, ICP with disqualifiers, job to be done, three named alternatives and proof points once, into .claude/context/positioning.md, so the deck, site, docs and launch post say the same thing. Use for "positioning", "ICP", "who is this for", "value prop", "how do we describe this".

Open .md

Just this skill, with its reference files, at the pinned commit:

$ mkdir -p .claude/skills/positioning && curl -fsSL https://codeload.github.com/solanabr/startup-builder-skill/tar.gz/a738bc8f54fcc39d40647ca791a5af375025c72d | tar -xz -C .claude/skills/positioning --strip-components=3 startup-builder-skill-a738bc8f54fcc39d40647ca791a5af375025c72d/skills/positioning

Or the whole set. The skills link to each other's format files, so the set works best together:

$ npx skills add solanabr/startup-builder-skill

No install: tell your agent

Read https://kit.superteam.codes/s/startup-builder-skill/skills/positioning/SKILL.md and follow it for this task:

SKILL.md

Open .md

Positioning

Pin down who it is for → what it replaces → why you win → write .claude/context/positioning.md. The artifact matters more than the interview: every other skill reads it instead of re-deriving the audience, and that is what stops the deck and the launch post from drifting apart.

Format and section rules: positioning-md-format.md.

Context handoff

  • At start: read .claude/context/positioning.md, .claude/context/idea.md and .claude/context/build.md if present. Take the wedge and the bear case from idea.md, and the proof points from build.md’s Traction table and What works today. Ask only for what is missing.
  • On completion: write .claude/context/positioning.md. Readers: pitch-deck, hackathon, idea-sprint (on a re-run).

Workflow

1. ICP, with who it is not for

Name the buyer and the user separately; on Solana they are often different. The ICP is often another team (a protocol integrating you, a wallet, another founder), and the one who decides may be a DAO or a foundation, not a person with a budget. Write disqualifiers, the people who will show up and are not the target. Crypto-specific patterns: crypto-icp.md.

2. Job to be done

One sentence: when , they want to , so they can . Use the situation they are in today, not the category you are in.

3. Three named alternatives

The real alternatives: a named protocol, the manual workaround, and doing nothing (or the spreadsheet, the Discord bot, the CEX). For each, the one dimension you beat it on, and where it beats you. Use real names; “existing solutions” is not an alternative. Mapping method and moat types: alternatives.md.

4. One-liner

Says what it does for whom, not what it is. Draft three:

  • a plain one a non-crypto friend understands (hackathon’s tagline test)
  • an “X for Y” anchor on something the ICP already respects
  • the builder version, with the jargon the ICP uses

Pick one as primary. A one-liner that still works with a competitor’s name swapped in is too generic.

5. Proof points with sources

Each one is a fact with a link: a tx, a build.md Traction row, an audit, a verified build, a named integrator. If a trust property is the differentiator (non-custodial, verified build, no admin key, immutable program), state it as the risk it removes for the ICP and prove it with the on-chain fact. Never claim a property that build.md’s upgrade authority contradicts.

6. Words to avoid

List the words this team must not use, with the replacement. Start from the defaults in the format file and add the ones this category overuses.

7. Write and check

Write positioning.md, then read the one-liner against the ICP and the disqualifiers: would a disqualified reader think it is for them? If so, tighten it.

Not here

  • Visual and verbal brand identity: get-shit-pretty’s gsp-brand-strategy can start from this file.
  • Anything about how a token, fee share or yield claim is classified, or what you may say about it where: crypto-legal-skill. This skill does not improvise there.
Solana AI Kit v2.4.0

One install for the whole kit

Superteam Brasil builds these skills as part of the Solana AI Kit, its open-source Claude Code and Codex kit for Solana. 15 specialist agents, 32 commands, 4 MCP servers and 46 skill packs, behind a firewall that keeps agents away from wallet keypairs.

Specialist agents
15
Commands
32
Skill packs
46
MCP servers
4
$ curl -fsSL https://aikit.superteam.codes | bash