# Design Goal

> Design a durable development goal that advances coherent autonomous work to the furthest useful user-review boundary. Use when the user asks to create, set, design, or choose a goal for ongoing repository work.

- Skill: `clairem0/design-goal` (Agent Skill)
- Install (CLI): `npx skillmds@latest add clairem0/design-goal`
- Raw SKILL.md: https://api.skillmd.com/api/skills/clairem0/design-goal/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: clairem0 (https://skillmd.com/u/clairem0)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/clairem0/design-goal

---


# Design Goal

Write one brief goal using five parts:

- **Outcome:** What should become true?
- **Authority:** Which plan, backlog, or source of truth governs scope? Point to it instead of repeating it. Authority defines scope; it does not grant permissions beyond the user's request.
- **Loop:** What coherent unit should the agent complete repeatedly?
- **Proof:** How should each unit be tested or reviewed from the end user's perspective?
- **Stop:** When should the agent finish, pause for user review, or ask for help?

Use this template, adapting or omitting clauses that add no value:

> Achieve **[outcome]**, guided by **[authority]**, using **[workflow loop]**. After each **[work unit]**, verify **[proof]** and resolve findings. Continue while **[safe conditions]** are met; stop at **[user decision, blocker, or completion]**.

Prefer the furthest useful review boundary that does not hide a decision the user must make earlier. If the environment supports durable goals and the user asked to set one, set it; otherwise return the goal text.