Skip to main content
Generalvudovn

verify-changes

Prove code works by running it, not just checking it exists. Verification through execution rather than inspection. Use after writing or modifying code to ensure it actually functions correctly.

Stars
7,619
Source
vudovn/antigravity-kit
Updated
2026-05-29
Slug
vudovn--antigravity-kit--verify-changes
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/vudovn/antigravity-kit/HEAD/.agents/skills/verify-changes/SKILL.md -o .claude/skills/verify-changes.md

Drops the SKILL.md into .claude/skills/verify-changes.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

Verify Changes — Prove Code Works

"Code that exists" ≠ "Code that works." This skill ensures changes are verified through execution.

Core Principle

❌ Verification by inspection:  "I can see the function exists, it should work"
❌ Verification by assumption:  "The types check out, so it's correct"
✅ Verification by execution:   "I ran it, here's the output, it works because [evidence]"

Verification Protocol

Step 1: Identify What Changed

- Which files were modified?
- What behavior should be different now?
- What was the original bug/requirement?

Step 2: Determine Verification Method

Change Type Verification Method
Bug fix Reproduce the original bug scenario → confirm it no longer occurs
New feature Run the feature → confirm expected output
Refactor Run existing tests → confirm nothing broke
API change Call the endpoint → confirm response shape
UI change Render the component → confirm visual output
Config change Load the config → confirm values applied
Build/infra Run build command → confirm success

Step 3: Execute Verification

# For Node.js projects
npm run build          # Does it compile?
npm run test           # Do tests pass?
npm run dev            # Does it start?

# For specific files
node -e "require('./path/to/module'); console.log('✅ Loads correctly')"

# For API endpoints
curl http://localhost:3000/api/endpoint

# For scripts
python script.py --test

Step 4: Report Evidence

## Verification Report

### What was changed
- [File list and summary]

### How it was verified
- [Exact commands run]

### Evidence
- Build: ✅ Compiled without errors
- Tests: ✅ 42/42 passing
- Runtime: ✅ Server starts, endpoint returns expected JSON
- Edge case: ✅ Empty input handled correctly

### Not yet verified
- [Anything that couldn't be tested automatically]

Verification Checklist by Project Type

Web Application

  • npm run build — compiles without errors
  • npm run lint — no linting errors
  • npm run test — all tests pass
  • Dev server starts successfully
  • Changed pages render correctly
  • No console errors in browser

API / Backend

  • Server starts without errors
  • Changed endpoints respond correctly
  • Error cases return appropriate status codes
  • Database queries execute successfully

CLI / Script

  • Script runs without errors
  • Expected output matches actual output
  • Error handling works (bad input test)
  • Help/usage text is correct

Anti-Patterns

Anti-Pattern Why It's Bad Fix
"It should work" No evidence Run it and show output
Only checking happy path Bugs hide in edge cases Test error paths too
Verifying only compilation Compiles ≠ correct Test runtime behavior
Skipping verification for "trivial" changes Trivial changes cause real bugs Verify everything

Integration with Other Skills

After Using Verify With
frontend-design → UI changes Render in browser, check console
backend-specialist → API changes curl endpoints, check responses
database-design → Schema changes Run migrations, query data
testing-patterns → New tests Run test suite, check coverage