prototype-first
Skill1 starsChanged 4 months agoMITView on GitHub
For visual, design, or "I'll know it when I see it" tasks, generate cheap throwaway prototypes BEFORE touching real code. Use whenever the user asks for UI, dashboards, design directions, video/visual output, landing pages, or any task where their taste cannot be specified in words. Also invocable as /prototype-first <task description>.
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
Converge once, then implement: After feedback, produce ONE refined prototype.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Prototype First
Purpose: the user's aesthetic and product taste is an unknown known — they have it, but cannot write it down. Words are a lossy channel for taste. Prototypes are a high-bandwidth channel. Spend cheap tokens on throwaway artifacts to extract taste before spending expensive tokens on real implementation.
Procedure
- Confirm this is a taste-driven task. Signals: the request contains words like "nice", "clean", "modern", "premium", or the deliverable is something the user will judge by looking at it. If the task has an objective spec (an API contract, a data transform), this skill does not apply — skip it.
- Generate divergent variations, not increments. Produce ONE self-contained HTML file containing 3–4 genuinely different directions side by side (or navigable via tabs). "Different" means different layout philosophy, density, typography, and mood — not the same design with four accent colors. Each variation gets a short label and one sentence on what it optimizes for.
- Use realistic data. Prototypes with lorem ipsum and fake numbers hide layout problems. Pull real or plausible data shapes from the actual project when available.
- Ask for a reaction, not a selection. Prompt the user: "Which parts of which variation feel right? You can mix." The goal is extracting taste vectors, not picking a winner.
- Converge once, then implement. After feedback, produce ONE refined prototype. When approved, implement it in the real codebase — and only then. Keep the prototype file around as the visual spec; reference it in IMPLEMENTATION_NOTES.md.
For non-UI visual work (video, images, documents)
Same principle, adapted medium: for video, render a short low-quality proof-of-concept segment before the full pipeline; for documents/reports, produce a one-page structural mock before the full document.
Anti-patterns
- Building variation 1 inside the real codebase "to save time" — it contaminates the code with an unvalidated direction and makes the other variations feel expensive.
- Four near-identical variations. If the user can't tell them apart at a glance, the pass was wasted.
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- K-9Nine/skipthespec★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- blindspot-passSkill
- convergence-guardSkill
- fable-methodSkill
- implementation-notesSkill
- interview-meSkill
- merge-quizSkill
- model-economySkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.