clarification-expert
Clarify ambiguous requests by researching first, then asking only judgment calls; stop before implementation.
When & Why to Use This Skill
The Clarification Expert is a specialized Claude skill designed to eliminate ambiguity and define clear success criteria for complex tasks. By prioritizing autonomous research and identifying missing requirements before implementation, it ensures that project goals are perfectly aligned with user expectations, significantly reducing wasted effort and technical debt.
Use Cases
- Initial Project Scoping: Transforming vague prompts like 'build a system' or 'optimize this' into concrete, actionable requirements with defined success criteria.
- Technical Discovery: Automatically researching existing codebases or documentation to resolve factual uncertainties, ensuring the user is only consulted for high-level judgment calls.
- Requirement Conflict Resolution: Identifying and highlighting contradictory instructions or implicit trade-offs in a request to ensure a consistent project direction.
- Pre-Implementation Planning: Generating a structured snapshot of facts and decisions to provide a solid foundation for developers or agents before they begin coding.
| name | clarification-expert |
|---|---|
| description | Clarify ambiguous requests by researching first, then asking only judgment calls; stop before implementation. |
Clarification Expert
When to use
- The request is ambiguous, under-specified, or missing success criteria.
- The user asks to “build a system”, “optimize”, “make it better”, or “how do I”.
- Requirements conflict, or trade-offs are implicit.
Quick start
- Research first; don’t ask for discoverable facts.
- Maintain a running snapshot (facts, decisions, open questions).
- Ask only judgment calls, in a numbered block.
- Incorporate answers and repeat until no open questions remain.
- Generate verbose beads, then stop (no implementation).
Snapshot template
Snapshot
- Facts:
- Decisions:
- Open questions:
Human input block
Use this exact heading and numbered list:
CLARIFICATION EXPERT: HUMAN INPUT REQUIRED
1. ...
2. ...
3. ...
Guardrails
- Never ask what the code can reveal; inspect the repo first.
- Keep questions minimal and sequential.
- After bead creation, hard-stop.
Deliverable format
- Snapshot.
- Human input block.
- One-line Insights/Next Steps.
Activation cues
- "clarify"
- "ambiguous"
- "build a system"
- "make it better"
- "optimize this"
- "how do I"
- "unclear goal"
- "conflicting requirements"