# Shape

> Clarify a complex or ambiguous request before execution by establishing the real problem, outcome, audience, scope, constraints, current state, and owner. Use when acting immediately risks solving the wrong problem.

- Skill: `aylee/shape-2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add aylee/shape-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/aylee/shape-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: aylee (https://skillmd.com/u/aylee)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/aylee/shape-2

---

<!-- Generated from .agents/skills/shape/SKILL.md; source-sha256: 8de540219afbaa4ce34dfb79a1879004638c4b622647367421c91c1422189f58. -->


# Shape work

Remain read-only while shaping.

1. Inspect discoverable repository and binder context before asking questions.
2. State the problem in plain language, why it matters, the desired outcome, audience, constraints, and known current state.
3. Separate facts, assumptions, preferences, and unresolved decisions.
4. Define “done” and name the owner-of-record artifact or workstream.
5. Recommend the smallest appropriate next move: answer directly, research, `$grill`, draft an artifact, plan implementation, or execute.

Ask only high-impact questions that cannot be answered from available context. If multiple interdependent decisions remain, route to `$grill`. Do not create durable state unless the user asks.

