AgentSkillsCN

Lab Experimentation

实验室实验

SKILL.md

Skill: lab-experimentation

Scope

Use lab experiments to answer "how should we do this?" questions about jsgui3 behavior.

Does:

  • •Run existing lab experiments to confirm behavior.
  • •Create minimal new experiments when behavior is unknown.
  • •Promote stable findings into Skills/Patterns.

Does Not:

  • •Replace unit tests (labs are for exploration).
  • •Cover production deployment scenarios.

Lab Structure

code
lab/
├── experiments/           # Individual experiments
│   └── 001-topic-name/
│       ├── README.md      # Hypothesis + expected result
│       ├── check.js       # Deterministic assertions
│       └── index.html     # Optional visual demo
├── fixtures/              # Shared test data
├── results/               # Captured outputs
└── README.md              # Lab index

Procedure

1. Check Existing Labs

bash
# List experiments
dir lab\experiments

# Read an experiment's README
type lab\experiments\001-example\README.md

2. Run an Experiment

bash
# Run check.js
node lab/experiments/001-example/check.js

3. Create New Experiment

bash
# Create folder
mkdir lab\experiments\NNN-short-slug

# Add files:
# - README.md (hypothesis)
# - check.js (assertions)

README.md Template:

markdown
# Experiment NNN: <Title>

## Hypothesis
<What you expect to happen>

## Setup
<Steps to run>

## Expected Result
<What success looks like>

## Actual Result
<Filled after running>

check.js Template:

javascript
'use strict';
const assert = require('assert');

// Setup
// ...

// Test
// ...

// Assertions
assert.strictEqual(actual, expected, 'Description');

console.log('✅ Experiment passed');
process.exit(0);

4. Distill Findings

  • •Update relevant Skill if pattern discovered.
  • •Add to docs/agi/PATTERNS.md if broadly reusable.
  • •Record in session working notes.

Validation

  • •Experiment check.js exits with code 0.
  • •README.md documents hypothesis and result.

References