Aftercheck — Testing skill for Claude Code

aftercheck

**Stop manually re-testing your whole app after every change.**

`aftercheck` is an [Agent Skill](https://github.com/rominirani/antigravity-skills) (the open `SKILL.md` standard) that analyzes what your recent code change might have affected — then actually *verifies* the affected areas with dummy data instead of just telling you what to go click on yourself.

Built for teams without an automated test suite, where "did I break anything else?" is normally answered by re-testing everything by hand.

you: "finished the order refactor, check what might break"

aftercheck:
  1. reads your git diff
  2. traces which other files reference what you changed
     (including DI/interface usage in C#, not just direct calls)
  3. actually calls the affected endpoints with fake data
  4. actually loads the affected pages and checks for errors
  5. compares screenshots against a visual baseline
  6. gives you a pass/fail report with evidence — not guesses

Why

Most "AI coding assistant" workflows stop at *suggesting* what to test. `aftercheck` goes one step further: it runs the checks itself, with clearly-labeled dummy data, and only reports something as "working" if it actually observed that with a trace to back it up.

It also remembers. Every run is logged, so files that have caused regressions before get flagged as high-risk automatically in future runs — even for small-looking changes.

Features

• 🔍 Impact analysis — traces changed code to everything that references it, including ASP.NET dependency-injection patterns (interface-based calls, not just direct class references) and Next.js file-based routing.
• 🎯 Confidence scoring — flags generic/noisy matches (low confidence) separately from specific, trustworthy ones (high confidence), so you're not drowning in false positives.
• ✅ Real verification, not just a checklist — sends actual dummy requests to local backend endpoints and actually loads affected frontend pages, rather th