Builds throwaway prototype experiments to validate technical feasibility before committing to a real build.
npx skills add https://github.com/ewtodd/son-of-anton --skill spike-ewtodd
Please help me install this Agent Skill. Skill: spike Source: https://github.com/ewtodd/son-of-anton/tree/main/skills/software-development/spike Command: npx skills add https://github.com/ewtodd/son-of-anton --skill spike-ewtodd
spikes/NNN-name/ directory. - Comparison spikes and verdicts: Runs parallel head-to-head variants (e.g., pdfjs vs camelot) with delegated subagents, then records a VALIDATED / PARTIAL / INVALIDATED verdict with surprises and recommendations in each spike's README. - Use Case: Before building a streaming chat feature, spike whether websocket token streaming reaches the client under 100ms, and compare two PDF parsing libraries for your ingestion pipeline — then throw the code away and plan the real build. ## Quick Start Ask the agent to spike whether your chosen approach works, for example: "Spike whether websockets can stream LLM tokens to the browser with under 100ms latency before I commit to this architecture."High-intent search queries and answers about installing and using this skill.
Run a spike: decompose the idea into 2-5 independent feasibility questions framed as Given/When/Then, order them by risk, then build a minimal runnable prototype for each. Close each spike with a VALIDATED, PARTIAL, or INVALIDATED verdict backed by observed evidence.
Build comparison spikes: give both variants a shared number with letter suffixes (002a, 002b), build them back to back or in parallel via delegated subagents, then write a head-to-head table covering quality, setup complexity, and performance before naming a winner.
Skip spiking when the answer is knowable from documentation or reading code, when the work is on the production path (use a planning workflow instead), or when the idea is already validated and you can jump straight to implementation.
A good spike question targets a specific feasibility claim with observable output, such as whether websocket streaming delivers chunks under 100ms. Avoid questions that are too broad, produce no observable output, or amount to just reading documentation.
Each spike lives in its own directory like spikes/001-websocket-streaming/ with a README and minimal code, biased toward something interactive such as a CLI, HTML page, or single endpoint. The README closes with a verdict section covering what worked, what didn't, surprises, and recommendations.
Spike code is throwaway by design: it hardcodes configuration, skips build tools, Docker, and env management, and is not production-ready. A spike that requires days of cleanup to ship was scoped incorrectly.
open-platform-modelGuide problem exploration and requirement clarification with clarifying questions and ASCII diagrams.
matthew-johnsonDecompose user ideas into independent feasibility spikes with given/when/then verdicts.
KbediakoSpawn parallel subagents for research, review, and planning tasks.
swcstudiospaceBuild throwaway prototypes to validate technical feasibility before committing to a real build.