prototype

Runs time-boxed spikes and throwaway prototypes to answer technical feasibility questions.

15|3|Updated Jul 9, 2026
One-click install
npx skills add https://github.com/kiurakku/cursor-kit-for-ai --skill prototype-kiurakku
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prototype
Source: https://github.com/kiurakku/cursor-kit-for-ai/tree/main/plugins/engineering/skills/prototype
Command: npx skills add https://github.com/kiurakku/cursor-kit-for-ai --skill prototype-kiurakku

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Teams often commit to specs and implementations before knowing whether an approach is technically feasible, wasting effort on dead ends. This Skill structures short, time-boxed spikes that answer one specific question with measurable evidence before any production code is written. ## Core Features & Use Cases - Spike Framing: Forces a one-sentence question, explicit time box, and success criteria before any code is written. - Throwaway Code Discipline: Defines branch naming, salvage rules, and deletion defaults so prototype code never silently becomes production code. - Structured Findings Report: Produces a verdict (proceed, pivot, blocked) with measured results, risks, and recommended next steps. - Use Case: Before committing to a CSV import feature, run a 3-hour spike to verify whether streaming parses 50k rows under 2 seconds, then hand the findings to the spec-writing workflow. ## Quick Start Ask the agent to run a time-boxed spike answering whether your chosen library or approach meets a specific measurable constraint, and produce a findings report.

Frequently Asked Questions about prototype

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a time-boxed technical spike before writing a spec?▼

State the spike question in one sentence, agree a time box (default 2-4 hours), and define success criteria that end the spike. Build the smallest experiment that can fail, measure results, then write a findings report with a proceed, pivot, or blocked verdict.

What makes a good spike question for a prototype?▼

A good spike question is specific and measurable, such as whether streaming parses 50k CSV rows under 2 seconds or whether a library supports a required OAuth flow. Vague goals like building a feature or trying some libraries are not valid spikes.

Should prototype code be merged into the main branch?▼

No. Spike code lives on a spike/<slug> branch, uses sandbox credentials, and is deleted or quarantined by default. If more than 20 percent would carry to production, stop and rewrite through the normal spec and implementation workflow.

When should I stop a spike early?▼

Stop when success criteria are met with margin, when a blocker requires a vendor or human decision, or when the approach is orders of magnitude wrong. Document what was learned and what remains unknown rather than silently extending the time box.

What is the difference between a spike and a production implementation?▼

A spike produces learning: latency numbers, API feasibility, and library fit, delivered as a findings report. Production implementation adds error handling, auth, tests, and observability, and only begins after the spike recommendation is converted into a spec.