Early-stage idea shaping: generate options, stress-test assumptions, then narrow to a testable direction. Distinct from planning-and-task-breakdown (execution) and spec-driven-development (requirements lock).
When to use
Idea is directional but not actionable
User wants alternatives before committing
Trade-offs between 2–3 approaches need surfacing
When not to use
Requirements are already agreed
User wants immediate implementation
Pure research with no decision deadline
Workflow
Phase 1 — Diverge (2–4 options)
Restate the problem in one sentence
Propose distinct approaches (not minor variants)
For each: upside, downside, effort tier (S/M/L), key assumption
Phase 2 — Stress test
What would make each option fail?
What data or spike would de-risk the choice?
What is explicitly out of scope?
Phase 3 — Converge
Recommend one direction with rationale
List what to validate in the next 1–2 hours/days
Present design in short sections the user can approve or correct before locking scope
Hand off to interview-me, spec-driven-development, or planning-and-task-breakdown
Before code
Do not start implementation during brainstorming
Do not auto-save specs or commit without explicit user request
Output template
## Idea refinement
- **Problem:**
- **Options considered:** (name + one-line summary each)
- **Recommendation:**
- **Why not the others:**
- **Next validation steps:**
Related skills
interview-me — when user intent is still unclear
doubt-driven-review — adversarial check before high-blast-radius choices
architecture-decision-records — when converged choice is architectural
1---2name: idea-refine-23description: Idea Refine4---56# Idea Refine78Early-stage idea shaping: generate options, stress-test assumptions, then narrow to a testable direction. Distinct from `planning-and-task-breakdown` (execution) and `spec-driven-development` (requirements lock).910## When to use1112- Idea is directional but not actionable13- User wants alternatives before committing14- Trade-offs between 2–3 approaches need surfacing1516## When not to use1718- Requirements are already agreed19- User wants immediate implementation20- Pure research with no decision deadline2122## Workflow2324### Phase 1 — Diverge (2–4 options)2526- Restate the problem in one sentence27- Propose distinct approaches (not minor variants)28- For each: upside, downside, effort tier (S/M/L), key assumption2930### Phase 2 — Stress test3132- What would make each option fail?33- What data or spike would de-risk the choice?34- What is explicitly out of scope?3536### Phase 3 — Converge3738- Recommend one direction with rationale39- List what to validate in the next 1–2 hours/days40- Present design in **short sections** the user can approve or correct before locking scope41- Hand off to `interview-me`, `spec-driven-development`, or `planning-and-task-breakdown`4243### Before code4445- Do not start implementation during brainstorming46- Do not auto-save specs or commit without explicit user request4748## Output template4950```markdown51## Idea refinement52- **Problem:**53- **Options considered:** (name + one-line summary each)54- **Recommendation:**55- **Why not the others:**56- **Next validation steps:**57```5859## Related skills6061- `interview-me` — when user intent is still unclear62- `doubt-driven-review` — adversarial check before high-blast-radius choices63- `architecture-decision-records` — when converged choice is architectural6465*Adapted from [addyosmani/agent-skills](https://github.com/addyosmani/agent-skills) and [obra/superpowers](https://github.com/obra/superpowers) (MIT).*
Run npx skillmds@latest add charlieviettq/idea-refine-2 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Idea Refine It is listed under Productivity on SkillMD.
This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
charlieviettq (@charlieviettq) published this skill. Their other Agent Skills are listed on their SkillMD profile.
charlieviettq