Vision grill

Vision Grill

What this skill does

vision-grill takes a vague product idea and interrogates it — adversarially, on strategy, positioning, market, and the product's reason to exist — until it is sharp enough to set a vision from. It produces a transient readied-idea stub that product-vision consumes. It is a pre-stage, not a pipeline stage: it writes no vision.md, no durable artifact, no why-link. It grills the product idea; product-vision authors the vision from the sharpened result.

When this skill applies

• A vague product idea that needs sharpening before a vision can be set — this skill.
• A product whose purpose, audience, and positioning are already clear — NOT this skill; go straight to product-vision. Grilling is the front door for vague product ideas, never a mandatory gate.
• A single feature idea — NOT this skill; that is idea-grill.
• Anything about how the product would be built, or marketing execution (campaigns, channels, copy) — NOT this skill.

The grill

Interrogate the product idea adversarially. Find the holes; do not be agreeable. Pursue, until each has a real answer:

• The problem space. What problem does this product exist to remove? Is it real and felt, or assumed? What do people do today instead?
• The audience. Who is this for — specifically? Who is it NOT for?
• Reason to exist. Why should this product exist at all? What changes for its users if it does? Why now?
• Positioning. How is this different from what people already use? What is the one-sentence statement of what it is?
• The boundary. What is this product deliberately NOT? Name the non-goals — the things it will refuse to become.
• Coherence. Do the answers above hang together as one product, or do they describe two?

Rules of the grill:

• Stay on strategy and desirability. Never ask about tech stack, architecture, or build effort. Never drift into marketing execution — campaigns, copy, channels are not this grill's concern; positioning and audience are.
• Push back. A grill that only agrees has failed. If an answer is thin, say so and ask again.
• Surface, do not resolve. When you find a contradiction or a weak point, NAME it and put it back to the product owner to resolve — do not hand them a reworded, cleaned-up version of their idea. Exposing the hole is your job; deciding how to close it is theirs. A grill that rewrites the product into a more defensible shape has authored its identity for the owner, however gently. Ask the sharp question and hold.
• Flag vision-level implications; do not resolve them. If grilling surfaces something that implies a change to the product's strategy or vision — a new way to make money, a new audience, a goal the product does not currently have — NAME it as a vision-level question and record it as an open item for the human to take to the vision. Do NOT decide it, and do NOT load or argue from the product's vision: your job is to grill the idea adversarially, not to defend the existing vision or adjudicate against it. Grilling may reveal the vision itself is wrong or incomplete — that is a finding, surfaced to the human, never settled inside the grill.
• Grill until the product idea is sharp — then stop.

The readied-idea stub

The readied-idea stub is defined in DESIGN_PRODUCT_PIPELINE.md §5.0. When the product idea is sharp, write one — NOT a document, and NOT a vision.

Location: a clearly-transient path — {declared-scratch-path} if the project declares one, else docs/.scratch/ — never the durable state tree. Filename: readied-product-{short-slug}.md.

Content — a handful of fields, no prose narrative, no vision goals:

• Product — one line: what it is.
• Problem space — the problem it removes, and for whom.
• Audience — who it is for, and who it is not.
• Reason to exist / why now — one or two lines.
• Positioning — how it differs from what exists.
• Deliberate non-goals — what the product refuses to be.
• Open strategy questions — anything the grill could not resolve.
• Vision-level implications — anything the grill surfaced that implies a change to the product's strategy or vision (new monetization, new audience, a goal the product lacks). Recorded for the human to take to product-vision — not resolved here.

product-vision authors vision.md from this stub and does not copy it (§5.0); once the vision exists, delete the stub. It must never grow into a vision — no vision goals, no why-link, no status. If you are writing vision goals, you have stopped grilling and started authoring — the wrong skill.

Handoff

End by telling the product owner the grill is done and the next step is product-vision, which will author vision.md from the readied-idea stub — eliciting any remaining intent, reconciling against existing capabilities, and resolving bootstrap tokens. Do not invoke product-vision yourself and do not pre-empt its work. Your deliverable is the sharpened product idea and the stub, nothing more.

Precedence

vision-grill writes nothing durable and overrides nothing. It reads no decision records and no rule files. If the grill surfaces a question about feasibility, implementation, or marketing execution, it does not answer it — it records strategy questions under "Open strategy questions" in the stub and drops the rest. The grill's authority is over the SHARPNESS of a product idea, never over what gets built or how it is taken to market.