Agent Skills: Design Exploration

>-

UncategorizedID: jagreehal/jagreehal-claude-skills/design-exploration

Install this agent skill to your local

pnpm dlx add-skill https://github.com/jagreehal/jagreehal-claude-skills/tree/HEAD/skills/design-exploration

Skill Files

Browse the full folder contents for design-exploration.

Download Skill

Loading file tree…

skills/design-exploration/SKILL.md

Skill Metadata

Name
design-exploration
Description
>-

Design Exploration

Critical rules

  • NO IMPLEMENTATION WITHOUT DESIGN AGREEMENT FIRST. No production code until the human validates the design.
  • One question per message; prefer multiple choice when options are known.
  • Present 2-3 approaches with trade-offs and a recommendation — never only one option.
  • YAGNI ruthlessly. Never silently fill ambiguous requirements.
  • Present design in 200-300 word sections; confirm each before continuing.
  • Before questioning patterns or writing the doc, read references/process-and-template.md.

Workflow

  1. Understand context: project state, related code, existing patterns.
  2. Clarify intent: purpose, constraints, success criteria — one question at a time.
  3. Explore 2-3 approaches with explicit trade-offs; lead with recommendation.
  4. Present design in sections (architecture, components, data flow, errors, testing); confirm each.
  5. Write validated design to docs/plans/YYYY-MM-DD-<topic>-design.md, commit, ask ready for implementation (implementation-planning / git-worktrees).

Resources

Validation

  • [ ] Existing state and patterns understood
  • [ ] Ambiguities clarified one question at a time
  • [ ] 2-3 approaches with trade-offs and recommendation
  • [ ] Design covers architecture, components, data flow, errors, testing
  • [ ] Human approved each section; doc saved and committed

Constraints

  • Skip for unambiguous single-file changes, typos, or already-approved designs.
  • Related: implementation-planning, git-worktrees, tdd-workflow.