Claude Code Plugins

Community-maintained marketplace

Feedback

jsgui3-lab-experimentation

@metabench/copilot-dl-news
0
0

Use repo lab experiments (src/ui/lab) to answer jsgui3 “how should we do this?” questions, then distill the validated result into durable docs + Skills.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name jsgui3-lab-experimentation
description Use repo lab experiments (src/ui/lab) to answer jsgui3 “how should we do this?” questions, then distill the validated result into durable docs + Skills.

jsgui3 Lab Experimentation

Scope

  • Run existing lab experiments to confirm behavior (event delegation, activation, MVVM, mixins, helpers)
  • Create minimal new lab experiments when encountering unknown or disputed jsgui3 behavior
  • Promote stable findings into the repo memory system (Skills + Patterns/Lessons + guides when appropriate)

Inputs

  • The question you’re trying to answer (1 sentence)
  • The target surface (control/server route/activation path/event semantics/perf)
  • Whether you need SSR-only, client activation, or browser semantics (delegation/bubbling)

Procedure

  1. Inventory prior art

    • Search docs first: node tools/dev/md-scan.js --guide --search "<topic>" --json
    • Then search sessions: node tools/dev/md-scan.js --dir docs/sessions --search "<topic>" --json
  2. Check whether a lab experiment already covers it

    • Open lab index: src/ui/lab/README.md
    • Open manifest: src/ui/lab/manifest.json
    • Optional: render the Lab Console card list (quick sanity): node src/ui/lab/checks/labConsole.check.js
  3. Run the smallest relevant check(s)

    • Each experiment should have a check.js (or check.all.js) that exits cleanly.
    • Prefer targeted runs, e.g.:
      • node src/ui/lab/experiments/002-platform-helpers/check.js
      • node src/ui/lab/experiments/004-theme-mixin/check.js
  4. If the answer is still unclear, create a minimal new experiment

    • Create src/ui/lab/experiments/NNN-short-slug/
    • Add README.md (hypothesis + expected result) and check.js (deterministic assertions)
    • Add an entry to src/ui/lab/manifest.json and the table in src/ui/lab/README.md
  5. If the behavior touches event semantics, run the delegation suite

    • Use the shared runner (reuses one browser/page):
      • node src/ui/lab/experiments/run-delegation-suite.js --scenario=005,006
    • Add/adjust scenarios if your change affects capture/bubble/selector matching/stopPropagation.
  6. If the behavior requires SSR + client activation + real interactivity, run a Puppeteer scenario suite

    • Use the scenario suite runner (single browser, many scenarios):
      • node tools/dev/ui-scenario-suite.js --suite=scripts/ui/scenarios/control-harness-counter.suite.js
    • This is the recommended fast path for validating control activation wiring (DOM linking, event listeners, state updates) without paying the cost of full Jest/Puppeteer E2E per test.
    • Prefer this over adding more “JS-only DOM semantics” scenarios when what you really need is jsgui3 activation behavior.
  7. Distill the validated finding into durable memory

    • Create/extend a Skill (this registry): docs/agi/SKILLS.md
    • Add/extend a Pattern if it’s broadly reusable: docs/agi/PATTERNS.md
    • Write a guide if it’s a subsystem-level discovery: docs/guides/…
    • Record evidence + commands in the current session directory under docs/sessions/...

Validation

  • Lab console renders: node src/ui/lab/checks/labConsole.check.js

  • Run the specific experiment check(s) you relied on (examples):

    • node src/ui/lab/experiments/001-color-palette/check.js
    • node src/ui/lab/experiments/002-platform-helpers/check.js
    • node src/ui/lab/experiments/004-theme-mixin/check.js
  • If delegation/event semantics are involved:

    • node src/ui/lab/experiments/run-delegation-suite.js --scenario=005,006,007,008,011,014
  • If you need SSR + activation + browser interaction coverage:

    • node tools/dev/ui-scenario-suite.js --suite=scripts/ui/scenarios/control-harness-counter.suite.js

Escalation / Research request

Ask for dedicated research if:

  • A behavior only reproduces with SSR + activation + bundling and no existing lab harness covers it
  • A change appears to require new shared harness utilities (browserHarness / activation harness)
  • The experiment needs performance instrumentation (counts, timings, memory) beyond a simple check

References

  • Lab index: src/ui/lab/README.md
  • Lab manifest: src/ui/lab/manifest.json
  • Delegation suite runner: src/ui/lab/experiments/run-delegation-suite.js
  • Puppeteer scenario suites guide: docs/guides/PUPPETEER_SCENARIO_SUITES.md
  • Pattern: “Lab-First jsgui3 Knowledge Gap Closure” in docs/agi/PATTERNS.md
  • General experimental research SOP: docs/agi/skills/experimental-research-metacognition/SKILL.md