| name | align |
| description | Ask targeted clarifying questions before executing a task so the agent and user align on scope, constraints, success criteria, and output format. |
Align Skill
Goal
Prevent wasted work by clarifying intent before execution.
Use when the user says:
/align
/align 5
ask me clarifying questions
make sure you understand
align before doing this
Question count
If the user provides a number, ask exactly that many questions.
Examples:
/align 5 -> ask 5 questions
/align three -> ask 3 questions
If no number is provided, ask 5 questions by default.
Question design
Each question must be:
- Decision-relevant.
- Non-obvious.
- Answerable quickly.
- Useful for narrowing the output.
- Accompanied by lettered options when possible.
Output format
## Alignment questions
1. What is the intended scope?
A. Minimal prototype
B. Production-ready implementation
C. Research/design only
D. Other: ...
2. What should I optimize for?
A. Speed
B. Correctness
C. Maintainability
D. Cost
E. Other: ...
3. ...
Answer instruction
End with:
Reply with compact answers like: 1B, 2C, 3A, plus any notes.
When to avoid asking
Do not use align if:
- The user asked for immediate execution.
- The task is trivial.
- The required constraints are already clear.
- The user explicitly says not to ask questions.
After receiving answers
- Restate the interpreted brief.
- Mention any remaining assumptions.
- Execute without asking more unless blocked.
Guardrails
- Do not ask generic filler questions.
- Do not ask questions whose answers would not change the work.
- Prefer options that expose tradeoffs.