Shape
Shaped work sits at the right abstraction: concrete enough to walk through, abstract enough to leave room. Appetite is a constraint you choose, not an estimate you compute. A shaped pitch has five ingredients — problem, appetite, solution, rabbit holes, no-gos — drawn at fat-marker altitude: a breadboard or rough sketch, never a wireframe, never a slogan.
Modes [LOAD-BEARING]
Mode-selection (hybrid)
Auto-detect from the user's phrasing, with slash-arg override:
- Raw idea,
shape this,pitch this,what's the appetite→ build-shape. gut check,vibe check,does this feel right→ shape-check.- Existing pitch plus
fix,is this well shaped,reshape→ to-good-shape. - Finished work plus
did we ship the bet,match the shape, results review → feel-shape. - Anything else → build-shape (default).
- Explicit override:
/shape build-shape | shape-check | to-good-shape | feel-shape. The override always wins.
build-shape
Shape raw work into a pitch:
- Set the appetite. Pick small batch or big batch. The appetite bounds the solution; a solution that exceeds it gets cut, the appetite stands.
- Rough the solution. Draw a breadboard (places, affordances, connections) or a fat-marker sketch. Load
references/breadboarding.mdfor the notation and altitude tests before drawing. - Hunt rabbit holes. Walk the solution end to end; each hole is declared solved-in-principle (state how) or patched out with a stated decision.
- Write no-gos. Name what this bet deliberately excludes.
Completion criterion: the pitch has all five ingredients, and the solution sits at fat-marker altitude — a wireframe or task list is over-shaped; a slogan is under-shaped.
shape-check
Interactive gut check of a pitch with the user, via the AskUserQuestion tool. Question shape:
- One single-select question per axis; answers to one axis never hide inside another.
- The
(Recommended)option is first and carries the default; picking it accepts the default. - At most 4 questions per fire; more axes go in sequential batches, dependency order.
multiSelectonly for additive picks (optional sub-scopes), never for axis-with-default semantics.
The axes come from the pitch's own ingredients: appetite right-sized? which scope cuts? each unresolved rabbit hole — patch or re-shape? no-go boundaries holding?
Completion criterion: every answered axis is folded back into the pitch; unanswered axes are listed as open bets — none silently dropped.
to-good-shape
Repair a badly shaped pitch. Diagnose first, one line:
- Over-shaped — design already done (wireframes, field lists, task tickets). Raise the altitude: redraw as a breadboard (notation in
references/breadboarding.md), discard the pixel decisions. - Under-shaped — words without a walkthrough, unbounded appetite. Force an appetite and walk one concrete path through the solution.
- Missing ingredients — add the absent ones; the other four constrain what the new one can say.
Then rewrite the pitch.
Completion criterion: the rewritten pitch passes the build-shape bar (five ingredients, fat-marker altitude); the diagnosis is stated in one line.
feel-shape
Results-shape-check: compare a finished artifact to the shaped bet, ingredient by ingredient.
- Problem: does the result address the shaped problem?
- Appetite: bet vs actual spend.
- Solution: does the built thing follow the breadboard's places and connections (terms defined in
references/breadboarding.md)? - Rabbit holes: which ones bit, and what did they cost?
- No-gos: respected or crossed?
Verdict, exactly one: shipped-the-bet | scope-crept | under-delivered | different-bet.
Completion criterion: every ingredient has a matched/violated line with evidence, plus the one top-line verdict.