Skip to content
Skills Library
Programming · Research

spike

/spike

Answer one open technical question with a small disposable experiment before real work starts. Use when feasibility, a library's behaviour, performance or a choice between approaches cannot be settled by reading docs or code.

Repository
dusk_code ↗
Licence
MIT
Last repository change
13.11.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 13.11.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Answer one open technical question with a small disposable experiment before real work starts. Use when feasibility, a library's behaviour, performance or a choice between approaches cannot be settled by reading docs or code.

Original author description

Answer one open technical question with a small disposable experiment before real work starts. Use when feasibility, a library's behaviour, performance or a choice between approaches cannot be settled by reading docs or code.

How it works

  1. Decide that a spike is needed
  2. Frame the question
  3. Build the smallest thing that answers it
  4. Report
Original SKILL.md headings · 56 lines, 412 words
SKILL.md preview
---
name: spike
description: Answer one open technical question with a small disposable experiment before real work starts. Use when feasibility, a library's behaviour, performance or a choice between approaches cannot be settled by reading docs or code.
source: "dusk_code"
author: "Neil Ermak"
license: MIT
---

# Spike

A spike buys knowledge, not code. It ends with an answer the user can act on and a
folder that can be deleted without loss.

## Decide that a spike is needed

Read first. If the docs, the source or the existing tests answer the question, report
that and stop. Run a spike only when the answer depends on behaviour that has to be
observed: an API under real input, a crate on this platform, a latency budget, two
designs compared on the same data.

Production work is not a spike. When the user wants the feature itself, plan it with
`tasks` and build it properly.

## Frame the question

Write down, before any code:

- **Question**: one sentence that can be answered yes, no or with a number.
- **Pass criterion**: what result settles it, for example "p95 under 20 ms for
  10k rows" or "the parser accepts every file in `fixtures/`".
- **Budget**: how long the spike may take. Stop at the budget and report what is known.
- **Out of scope**: error handling, configuration, style, tests beyond the check itself.

Show this frame to the user when the question came from them loosely; a sharp question
halves the work.

## Build the smallest thing that answers it

- Work in a throwaway folder in the thread's workspace, or a scratch branch, never in
  the code that ships.
- Hard-code inputs. Print results. One file is usually enough.
- Use real data or a faithful sample: a spike on toy input answers a toy question.
- To compare approaches, run them on the same input, in the same process or harness,
  several times, and keep the raw numbers.
- Run it with `run_command` and read the actual output; never infer a result from code
  that was not executed.

Independent variants can run as separate agents with `ask_agent`, each with the frame
and one variant, when they would otherwise take long one after another.

## Report

Answer in this order:

1. The answer to the question, and whether the pass criterion was met.
2. The evidence: the command, the input and the measured output.
3. What the spike did not cover and could still change the answer.
4. A recommendation for the real implementation, including what to keep from the spike
   (an API call sequence, a configuration value) and what to discard.

Then remove the throwaway folder or say where it is, so nobody mistakes it for working
code.

Versions

RepositoryScoreLicence
This page · DroBov1Tya/dusk_code100MIT
hashintel/labs100MIT
kjgarza/marketplace-claude100MIT
cryndoc/polisade-orchestrator100Apache-2.0
srstomp/pokayokay100MIT
uta2000/feature-flow100MIT
JoDi-27/JoDi-SDD90MIT
juanchaopan/dev-workflow90MIT
petekp/claude-code-setup81MIT

Database snapshot: 13.11.2026 · package version 8b873134903e (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100prototyping
    Answer ONE named empirical question with throwaway code — a disposable spike, never the shipped construction of `build`/`build-ui`. Use it when a question can only be settled by running something — trigger on \"spike this\", \"prototype it\", \"just try it and see\", \"build a…
  • 90prototype-spike
    Build or plan an explicitly throwaway prototype to answer a design, API, UX, performance, or integration question, then capture findings and cleanup. Use when the user asks for a spike, prototype, proof of concept, or experiment before committing to production design.
  • 100tech-spike
    Runs a time-boxed technical spike with the prototyper to answer one risky question (can this API do X, will this library work on the edge runtime, is this fast enough) and records the finding. Use when an estimate has low confidence, before an ADR, or when an integration is unproven.
  • 100spike-test-setup
    Configure spike test setup operations. Auto-activating skill for Performance Testing. Triggers on: spike test setup, spike test setup Part of the Performance Testing skill category. Use when writing or running tests. Trigger with phrases like "spike test setup", "spike setup", "spike".
  • 90throwaway-spike
    Use when a screen's look or a state model's soundness is unclear and disposable running code would settle it sooner than talk. Skip settled designs. Side-branch commit needs OK.
  • 100fai-technical-spike
    Run timeboxed technical spikes to de-risk unknowns with structured experiments, decision criteria, and actionable conclusions. Use when evaluating technologies, patterns, or approaches before committing to implementation.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.