andreaswasita/copilot-agents-dojo/demand-elegance
Claude Code and agent skill by andreaswasita/copilot-agents-dojo
- community
Checking that the SKILL.md file is still in the repository.
Challenges hacky fixes on non-trivial changes.
Install
npx vskill@latest install andreaswasita/copilot-agents-dojo/demand-eleganceSKILL.md
Demand Elegance Skill
Forces the agent to challenge its own non-trivial diff before presenting it: fewer moving parts, standard patterns over custom abstractions, boring over clever. Does NOT apply to typos, one-line fixes, or hotfixes — over-engineering trivial work is the failure mode this skill explicitly avoids.
When to Use
- A fix feels hacky, fragile, or "barely working".
- Multiple solutions exist and the tradeoffs aren't obvious.
- The diff is large and includes infrastructure the original task didn't ask for.
- Refactoring with simplification as the explicit goal.
- Self-review before presenting any non-trivial solution.
- NOT for: typos, one-line bugfixes, comment edits, formatting, production hotfires.
Prerequisites
- A working solution (or near-working) to challenge — don't run this on a blank canvas.
- The full diff visible (
git diffvia thepowershelltool). - Willingness to throw work away if a simpler approach exists.
How to Run
Quick Reference
| Question | If answer is "yes" → action | |
|---|---|---|
| Is there a more elegant way? | Try it before shipping. | |
| Knowing what I know now, would I build it this way? | If no, refactor before presenting. | |
| Can I reduce moving parts? | Cut them. | |
| Would a senior engineer raise an eyebrow? | They would. Fix it. |
| Choose | Over | |
|---|---|---|
| Fewer lines | More lines (all else equal) | |
| Standard patterns | Custom abstractions | |
| Fewer dependencies | More dependencies | |
| Readable | Clever | |
| Boring & correct | Exciting & fragile |
Procedure
Step 1: The Elegance Check
For non-trivial work, before presenting the diff, ask in order:
- Is there a more elegant way?
- Knowing everything I know now, would I build it this way?
- Can I reduce the number of moving parts?
- Would a senior reviewer raise an eyebrow at this?
A "yes" to any means: pause and try the simpler path.
Step 2: The Simplicity Ladder
When picking between solutions, prefer the upper rungs:
- No code change (the bug isn't a bug, or the requirement is wrong).
- Configuration change.
- Small, local code change inside one function.
- Localized refactor inside one module.
- Cross-module change with explicit motivation.
- New abstraction (only if 1–5 are insufficient).
If you land on rung 6, justify why 1–5 do not work in a comment or in tasks/todo.md.
Step 3: Self-Review the Diff
Run git diff via the powershell tool and read your own diff as if reviewing a stranger's PR. Identify the smelliest part — there is always one. Ask whether that smell can be eliminated, not just covered up.
Step 4: Restart If Hacky
If the diff still feels hacky after Step 3, throw it away and start over with the simpler approach. Resist the sunk-cost reflex.
A useful self-prompt: "Knowing everything I know now, implement the elegant solution."
Pitfalls
- DO NOT run this skill on trivial changes. A typo fix doesn't need an architecture review.
- DO NOT premature-abstract. Build for the requirements that exist, not the ones you imagine.
- DO NOT gold-plate. The user asked for a button fix, not a component-system redesign.
- DO NOT confuse cleverness with elegance. If you need a comment to explain what (not why) the code does, it's too clever.
- DO NOT apply during production hotfires. Stabilize first; log a
tasks/lessons.mdentry to revisit. - DO NOT deodorize a smell — eliminate it. Wrapping bad code in a clean interface still leaves bad code.
Verification
- [ ] Diff size is justified by the task description in
tasks/todo.md. - [ ] No new abstractions without explicit justification.
- [ ] No new dependencies without explicit justification.
- [ ] Self-review pass completed (
git diffread end-to-end). - [ ] No
tasks/lessons.mdentry taggedover-engineeringfor this task.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/demand-elegance/SKILL.md
Popularity
- Repository stars
- 32
- Repository forks
- 12
- 7-day trend score
- 17
Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 28 Apr 2026, 6 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/andreaswasita/copilot-agents-dojo/demand-elegance)