Shape Idea

Shape an Idea

Create shared understanding before exploring solutions. Keep the work at problem-and-scope level; do not design the system or plan its implementation.

Work with the user

• Inspect the conversation, repository, existing product behavior, and nearby documents before asking anything. Resolve accessible factual questions yourself and cite their sources in the brief.
• Ask one high-leverage question at a time. When asking for a choice, give a recommendation and its main tradeoff.
• Ask the user for judgments and priorities, not facts available in the working context. Never make a product or scope decision silently.
• Scale the depth to the idea. Preserve useful ambiguity while the idea is young, but make every uncertainty visible.
• Keep facts, assumptions, hypotheses, decisions, and open questions separate. Do not promote one category into another without new evidence or explicit agreement.

Classify the project context

Classify the work before shaping its boundaries:

• Greenfield: No implementation baseline exists. Treat proposed platforms, structure, and conventions as choices or assumptions rather than facts.
• Existing system: The idea changes a running product or established codebase. Ground the brief in current behavior, constraints, integration points, and behavior that must be preserved.
• Hybrid: A new component or product must connect to an existing system. State which side is new, which side is established, and where the seam lies.

Do not infer Greenfield merely because a repository is unavailable. Use Unclear when the evidence is insufficient and resolve the classification only when it could change scope, constraints, or the next decision. Carry the classification and its evidence into the brief.

Shape the brief

1. Ground the idea. Summarize the starting vision and inspect relevant context. Record what is known, where it came from, and what remains uncertain.
2. Find the problem. Identify who is affected, what happens today, what evidence supports the problem, why it matters now, and what a better outcome would change. Reframe a proposed feature as a problem when necessary.
3. Choose a meaningful slice. Define the smallest coherent scope that can create or test value. If the idea is a grab-bag, propose smaller slices and ask the user which one to shape.
4. Set appetite and boundaries. Capture the acceptable investment and risk, in-scope behavior, non-goals, constraints, and behavior that must be preserved.
5. Expose uncertainty. Record assumptions, falsifiable hypotheses, risks, rabbit holes, and open questions. Investigate cheap factual unknowns from available context; describe evidence needed for hypotheses.
6. Make success observable. State outcome signals that could show whether the idea helped without pretending that early targets are proven facts.
7. Write the artifact. Copy and complete assets/idea-brief-template.md. Remove unused guidance rather than inventing content.

Apply the readiness gate

Read references/readiness.md before declaring the brief ready.

• If a blocking item is missing, keep the status Draft, name the blocker, and ask only the next question needed to resolve it.
• If the gate passes, set the status to Ready for approval, present the complete brief, and ask the user to approve or revise it.
• After explicit approval, record the approval and set the status to Approved.

Hard stop: Do not explore solution options, write a design doc, or plan implementation until the user explicitly approves the written brief. After approval, recommend explore-options as the next stage without starting it automatically.