skills.rest
Wirasm
WirasmCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

prp-spike

Builds throwaway prototypes in isolated worktrees to prove or disprove technical hypotheses with evidence-backed verdicts.

2.2k|608|Updated Jun 21, 2025
Search Tags:#proof of concept#prototyping#feasibility#evidence#hypothesis testing#spike#Git worktree
One-click install
npx skills add https://github.com/Wirasm/PRPs-agentic-eng --skill prp-spike
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prp-spike
Source: https://github.com/Wirasm/PRPs-agentic-eng/tree/main/plugins/prp-core/skills/prp-spike
Command: npx skills add https://github.com/Wirasm/PRPs-agentic-eng --skill prp-spike

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Teams often debate feasibility, fit, or approach choices without evidence, wasting weeks on ideas that fail late. This Skill settles open technical questions by building the smallest artifact that could falsify a hypothesis, ending in a PROVEN, DISPROVEN, or CONDITIONAL verdict backed by reproducible evidence.

Core Features & Use Cases

  • Falsifiable Framing: Converts vague questions into precise hypotheses with pre-committed kill criteria and verdict boundaries before any code is written.
  • Isolated Experimentation: Runs spike code in a dedicated git worktree that never opens a PR, keeping disposable experiments away from production branches.
  • Evidence-Based Verdicts: Produces a structured spike report with observed evidence, verdict provenance, limits, and routing recommendations for the decision.
  • Use Case: When asked "can our schema support real-time collaboration without a migration", the Skill frames a falsifiable claim, builds a minimal probe against the real schema, stress-tests it, and delivers a verdict with the evidence stored for anyone to re-check.

Quick Start

Ask the agent to spike whether a specific idea is feasible, for example: spike whether we can stream large query results through our current API layer without hitting memory limits.

Frequently Asked Questions about prp-spike

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

FAQPage Schema
How do I test if a technical idea is feasible before committing to it?▼

Frame the idea as a falsifiable hypothesis with kill criteria, then build the smallest artifact that could prove it wrong. The spike ends in a PROVEN, DISPROVEN, or CONDITIONAL verdict backed by observed evidence, not a shippable implementation.

What is the difference between a spike and a prototype?▼

A spike answers one falsifiable question and its code is throwaway by construction, never opening a PR. A prototype aims toward a usable product. Spike code that looks good must still be rewritten under normal standards via plan-and-implement workflows.

How do I compare two implementation approaches fairly?▼

Fix the winning metric before building either variant, vary only one thing, give each approach its best implementation, and repeat measurements to report spread. A metric chosen after seeing results will favor whichever approach you preferred.

When should I not use a spike workflow?▼

Skip spiking when feasibility is already settled (use planning and implementation instead) or when something is already broken (use debugging). Also avoid spikes blocked on credentials or decisions only a human can make.

What does a CONDITIONAL verdict mean in a spike report?▼

CONDITIONAL means the hypothesis holds only if a named constraint changes, such as a primitive, schema, or product rule. The report prices that change, states its blast radius, and frames it as a decision rather than a task.

Related Skills

Explore More Software Engineering Skills→
swarm-ai-researchswarm-ai-research

Research Brief

Synthesize web and academic sources into a falsifiable research brief with citations.

Official
Advanced
rdleclercrdleclerc

uberplan

Create long-running engineering plans with scope fidelity and evidence-backed proof gates.

Community
Advanced
synqingsynqing
1

dispatching-parallel-agents

Dispatch parallel Claude agents to investigate and fix independent failures.

Community
Advanced
Lev-itLev-it

grill-with-docs

Analyze technical plans through interactive questioning and generate ADRs and glossaries.

Official
Intermediate