skills.rest
gitkraken
gitkrakenOFFICIAL
Explore all skills by this author →
Website𝕏@gitkrakenGitHub

live-exercise

Exercises running VS Code UI features through live inspection, audit lenses, and iterative fix loops.

9.9k|1.8k|Updated Aug 8, 2016
Search Tags:#ui testing#VS Code#Debugging#webview#ux audit#live inspection#Extension development
One-click install
npx skills add https://github.com/gitkraken/vscode-gitlens --skill live-exercise
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: live-exercise
Source: https://github.com/gitkraken/vscode-gitlens/tree/main/.claude/skills/live-exercise
Command: npx skills add https://github.com/gitkraken/vscode-gitlens --skill live-exercise

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Static code review cannot catch bugs that only appear when a UI actually runs — broken event handlers, layout overflow, missing loading states, or features whose behavior diverges from what users expect. This Skill operates a live VS Code instance to verify UI-bearing work end-to-end, from tactical bug fixes to ship-gate audits.

Core Features & Use Cases

  • Four-lens live auditing: Applies functional walk-through (L1), intent compliance via cold-read (L2), UX heuristics polish (L3), and opportunistic improvement hunting (L4) on every sweep of the running UI.
  • Iterative fix-and-verify loop: Compiles findings into structured docs, dispatches parallel fix agents, rebuilds, re-measures, and re-sweeps until convergence across simplify and performance gates.
  • Evidence-based verification: Uses the vscode-inspector MCP tools (evaluate_in_webview, aria_snapshot, read_console) to confirm handlers fired and state updated, rather than trusting diffs or type checks.
  • Use Case: After refactoring a GitLens webview panel, invoke the skill to launch VS Code, exercise every state of the panel, catch a missing loading spinner and a console error, dispatch fix agents, and verify the fixes live before shipping.

Quick Start

Ask the AI to run /live-exercise on the feature you just changed so it launches the extension, exercises the UI live, and fixes any issues it finds.

Frequently Asked Questions about live-exercise

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

FAQPage Schema
How do I verify a VS Code extension UI actually works before shipping?▼

Run /live-exercise to launch a live VS Code instance via the vscode-inspector MCP, exercise every feature state as a user would, and verify handlers fired using evaluate_in_webview. It audits functional behavior, intent compliance, and UX polish, then dispatches fix agents and re-verifies after rebuild.

What is the difference between live-exercise and static code review skills?▼

Static skills like /review and /deep-review analyze diffs and trace code paths without running anything. /live-exercise operates the running UI, measuring live state such as computed layout, console errors, and actual handler outcomes, catching bugs that only appear at runtime.

Does live-exercise require any setup or prerequisites?▼

Yes. The vscode-inspector MCP must be connected (auto-discovered via .mcp.json) and the build must currently pass with pnpm run build:quick. A goals.md file describing intended user experience is helpful but optional, since cold-read intent inference always runs.

When should I not use live-exercise for a UI change?▼

Skip the full loop for trivial changes with no state surface, such as single-line copy edits, color-token swaps, icon renames, or comment-only changes. Those earn only a build, diff, and one-click eyeball verification rather than findings docs and convergence checks.

Why does live-exercise insist on live verification instead of trusting TypeScript compilation?▼

Type checks and clean diffs only prove code is syntactically valid, not that it works for users. Handlers may not fire, state may not update, and layout may overflow at runtime. The skill treats phrases like 'tsc passed' as red flags requiring live evidence.

How does live-exercise handle fixes it finds during an audit?▼

Unambiguous findings are dispatched to parallel background fix agents (capped at 5-6), each with exact file paths, root causes, and measurable success criteria. After agents complete, the skill rebuilds, diffs the changes, resets instance state, and re-sweeps all lenses until convergence.

Related Skills

Explore More Software Engineering Skills→
dfrysingerdfrysinger
36

test

Run acceptance tests against implemented code and present pass/fail results.

Community
Advanced
gqf2008gqf2008

requesting-code-review

Automate pre-commit verification with static security scans and quality gates.

Community
Intermediate
trailofbitstrailofbits
6.5k

audit-prep-assistant

Coordinate audit readiness with static analysis, test coverage, and documentation artifacts.

Official
Advanced
tacogipstacogips
1

e2e-tdd

Guide UI feature development through the Plan-Red-Green-Refactor-Validate TDD cycle.

Community
Advanced