Goal Definition

Take a raw idea or request and turn it into a clear, actionable goal with explicit success criteria, assumptions, and constraints.

When to Use

• You have a vague idea or user request that needs sharpening
• Before brainstorming approaches — define the "what" before the "how"
• Requirements are ambiguous or incomplete
• You need to align on success criteria before investing in design
• Assumptions are lurking but haven't been stated

When NOT to use: The goal is already crystal clear with written success criteria, or a goal-definition.md already exists and is still accurate.

Input

• Raw idea or request from the user
• context-map.md from the artifact folder (recommended but not required)
• See references/context-sources.md for additional project documentation that may provide context

Output

• goal-definition.md saved to the artifact folder (see references/formats.md for template)

The Process

Step 1: Locate Artifact Folder

If context-map.md exists in a known artifact folder, use that same folder. Otherwise, ask the user for the artifact folder path (same as research Step 1).

Step 2: Restate as "How Might We"

Take the raw idea and reframe it as a crisp problem statement:

Raw: "We need better search" Reframed: "How might we help users find the content they need in under 3 seconds?"

This forces clarity on what's actually being solved. Present the reframed statement and confirm with the user before proceeding.

Step 3: Ask Sharpening Questions

Ask 3-5 targeted questions — no more. One question per message. Focus on:

1. Who is this for? — The specific user and their situation
2. What does success look like? — Concrete, observable outcomes
3. What are the real constraints? — Time, tech, resources, compatibility
4. What's been tried before? — Prior attempts, why they failed or didn't happen
5. Why now? — What changed that makes this important

Use structured options where possible:

What is the primary user for this feature?
  A. New users during onboarding
  B. Power users with large datasets
  C. All users equally
  D. Internal team / admin users

Do NOT proceed until you understand who this is for and what success looks like. These two answers shape everything downstream.

Step 4: Surface Assumptions

Before writing the goal document, list every assumption you're making:

ASSUMPTIONS I'M MAKING:
1. The existing search index can handle the new query patterns
2. Users primarily search by title, not content
3. We can add a new API endpoint without breaking existing clients
4. Response time under 500ms is achievable with current infrastructure
→ Correct me now or I'll proceed with these.

Wait for confirmation. Unvalidated assumptions are the #1 source of wasted work downstream.

Step 5: Reframe as Testable Success Criteria

Turn vague goals into concrete, testable conditions:

Vague: "Make search better" Testable:

• Search returns relevant results for 90% of common queries
• Results appear within 500ms
• Users can filter by date, type, and author
• Zero-results page suggests alternative queries

Present and confirm with the user.

Step 6: Define Scope Boundaries

State explicitly what is NOT included and why:

• Not Doing: [Thing] — [Reason]
• Not Doing: [Thing] — [Reason]

The "Not Doing" list is arguably the most valuable part. Focus is about saying no to good ideas.

Step 7: Write Goal Definition

Write goal-definition.md to the artifact folder using the template from references/formats.md. Include all findings from Steps 2-6.

Announce the saved path:

"Goal definition saved to [path]/goal-definition.md."

Common Rationalizations

Red Flags

• Jumping to solutions before defining the problem
• No "who is this for" answer
• Success criteria that say "works correctly" or "is fast"
• No assumptions listed
• Empty "Not Doing" list — everything has scope boundaries
• Proceeding without user confirmation on reframed problem

Verification

Before handing off, confirm:

• Problem restated as "How Might We" and confirmed by user
• 3-5 sharpening questions asked and answered
• Assumptions listed and confirmed
• Success criteria are specific and testable
• "Not Doing" list has at least 2 items
• goal-definition.md saved to artifact folder

Next

"Goal defined. Next recommended skill: brainstorming — explore approaches to achieve this goal."