Skill: verification-loop

Trigger: "verify" or "run verification loop" or "check this works" or after any code change

Auto-triggered by: debug-session after Step 5 (fix applied). Fires automatically — user does not need to ask.

Description: Continuous self-evaluation after code changes. Combines compile/build verification with functional smoke tests and a code quality self-check. Catches problems before the user has to report them.

Allowed Tools: Bash, Read, Glob

Steps

1. Build/Compile Check

• Run your project's build command
• Verify exit code 0 / no errors
• If build fails: fix before proceeding

2. Smoke Test

• Identify which endpoint(s) or function(s) were changed
• Call/invoke each with minimal valid inputs
• Verify expected output — not just "no error"
• Check for runtime errors in logs

3. Frontend Check (if applicable)

• Verify the changed UI element renders correctly
• Check that no JS errors would appear in console
• Verify no undefined values in rendered output

4. Self-Evaluation Checklist

• Does the change do exactly what was asked — no more, no less?
• Are edge cases handled? (empty values, null, missing params)
• Does it follow the project's conventions?
• Are memory files updated? (function indexes, reference docs)
• Any new code that introduces the same pattern we just fixed elsewhere?

5. Report

• ✅ "All checks passed — [summary of what was verified]"
• ❌ "[Issue found] — [description and suggested fix]"

Notes

• Run this skill after every substantive code change, before reporting done
• "It compiled" is not the same as "it works" — always smoke test
• Update memory/reference docs as part of verification, not as an afterthought
• If verification reveals a new issue, fix it before marking the task complete