Positioning
A 15-minute forced-choice session that produces a positioning statement and the artifacts a startup needs to hold a position: message architecture, mind virus, compass, ignore list, worldview, archetype, and a 90-day proof plan.
Position or be positioned. If you don't pick a category, you will be assigned one. The market's default setting is disbelief, so competition has shifted from making claims to surviving verification. This skill forces the choices, then makes every claim either provable today or a dated bet.
Operating rules (read first)
Questions. If an interactive question tool is available (for example AskUserQuestion in Claude Code or Cowork), use it for every forced choice. Otherwise ask in plain text: one question at a time, numbered options, wait for the answer. Never stack questions in plain-text mode.
Forced choice. No "all of the above". Gut answers over polished answers — tell the user 30 seconds per question is enough. The pain of choosing is the exercise.
Time. The session targets 15 minutes total. Never block on research; never loop. Phase budgets are in the session script. If research has not returned by the statement phase, proceed without it and mark the pack "research pending".
Research. If a subagent tool is available, launch research in the background during setup. If only web search is available, run a few quick searches inline between phases. If there is no web access, skip research and say so in the output. Never invent competitors.
Files. All artifacts live in positioning/ under the current working
directory (create it if needed). Never overwrite another founder's answers
file.
Which moment is this?
If the user asked for a specific mode in plain words ("compile", "start
over", "new session"), obey that. Otherwise check positioning/ in the
current directory and route:
| State | Moment | Script |
|---|---|---|
No positioning/ dir, or no answers-*.md files | First run — run the session | references/session.md |
One or more answers-*.md, no positioning.md | Ask: run your own session, or compile the team's answers now? | references/session.md or references/compile.md |
positioning.md exists | Offer: a new founder session, or recompile to update the pack | references/session.md or references/compile.md |
A solo founder goes straight from session to the full pack — compile is only for teams.
The method in one breath
Category first (compete / reframe / create — the maturity of the category decides which is honest). Then community (who you speak to, who you ignore). Then the logical spine: what is it, why does it matter, why you, why now. Then compress into an ONLYness or FIRSTness statement under 30 words, find the phrase the market could repeat without you in the room, pick the archetype that gives the position feel, and book three proofs for the next 90 days. Detail lives in the reference scripts — load only the one you need.
Quality rules (apply to every artifact)
- No sentence over 20 words. Statement under 30 words.
- No unverifiable adjectives: innovative, cutting-edge, world-class, unique, seamless, revolutionary.
- Every claim is either true today or explicitly a bet with a proof date.
- Aspirational is written "By [date] we will..." — never "We are...".
- If it reads like a SaaS landing page, rewrite it drier.
Files in this skill
references/session.md— the 15-minute founder session (phases, questions, timings)references/research.md— competitor research briefs and timeboxreferences/archetypes.md— the 90-second archetype pickerreferences/compile.md— merge co-founder answers, force tiebreaks, emit the packassets/pack-template.md— template forpositioning/positioning.mdassets/answers-template.md— template for per-founder answers files