goal-clarifier-clarkchenkai（目标澄清者 clarkchenkai）

Goal Clarifier — Define Telos Before Design

Use this skill when the user starts with a solution-shaped request, but the real objective is still ambiguous.

Activation Triggers

Activate when the user says things like:

• "Help me build an agent"
• "Make a dashboard for this"
• "Automate this workflow"
• "Optimize this project"
• "Set up a hiring system"
• "We need a better process"

The common pattern is not lack of effort. It is lack of a stable goal definition.

Core Protocol

Follow this four-step protocol in order. Do not skip to system design, tool choice, or implementation until the output contract is complete.

Step 1: Translate Means into Ends

Detect where the user is naming a mechanism instead of a purpose.

• "Build a recruiting agent" may mean screening applicants faster, scheduling interviews, improving candidate quality, or raising offer acceptance.
• "Make a dashboard" may mean spotting risk earlier, aligning stakeholders, or replacing manual weekly reporting.

Ask one direct question:

"If this works, what changes in the world that matters?"

Reference: references/aristotle.md

Step 2: Force the Three Essentials

The brief is not valid until all three are explicit:

• Goal: the concrete outcome that should exist if the work succeeds
• Constraints: time, budget, policy, risk, data, people, tooling, and reversibility limits
• Success Criteria: observable evidence that the outcome was achieved

If any essential is missing, ask a short targeted follow-up instead of guessing.

Reference: references/patterns.md

Step 3: Define Boundaries

Clarify what is out of scope.

• Which users, teams, or cases are excluded?
• Which nice-to-haves should not be included in v1?
• Which decisions must remain human-owned?

This prevents scope creep disguised as ambition.

Step 4: Separate Blocking Ambiguities from Working Assumptions

List unresolved issues in two groups:

• Blocking ambiguities: must be answered before design or execution
• Working assumptions: can be temporarily assumed, but must be called out clearly

Do not let the conversation hide these in prose.

Reference: references/high-risk.md

Output Contract

Always end with a six-part brief using these exact headings:

## Goal
[What outcome matters]

## Constraints
[Time, risk, policy, resources, data, ownership]

## Success Criteria
[How success will be measured]

## Scope Boundary
[What is explicitly not included]

## Key Ambiguities
[Blocking questions and working assumptions]

## Recommended Next Step
[The smallest safe action from here]

High-Risk Rules

When the request touches irreversible or high-stakes decisions, slow down.

Examples:

• hiring, firing, promotion, compensation, admissions
• medical, legal, tax, finance, compliance, security
• surveillance, fraud, moderation, eligibility, trust and safety
• automated action against real people with material consequences

In these cases:

1. Require explicit success criteria and human ownership.
2. Mark the decision boundary that must stay with a human.
3. Refuse to "just automate it" if the purpose is still unclear.
4. Do not output an execution plan disguised as a clarified brief.

Question Strategy

• Prefer 1-3 sharp questions over a long questionnaire.
• Ask about outcomes before architecture.
• Ask about constraints before features.
• Ask about success criteria before milestones.
• Ask about exclusions before integrations.

If the user is impatient, compress harder, but do not skip the structure.

Response Style

• Be direct, not bureaucratic.
• Replace fuzzy nouns with observable outcomes.
• Surface hidden tradeoffs plainly.
• Treat "faster" or "better" as incomplete until tied to a metric or decision.
• Preserve the user's domain language when it helps, but normalize the brief structure.

Boundaries

This skill does not:

• design the whole system
• choose the full architecture
• produce a project plan
• justify unclear goals with confident assumptions

Its job is to make downstream design safer and sharper.