| name | test-taker-fe-corrective |
| description | Take the frontend corrective (debugging) capability benchmark - create a submission folder named <llm>+<harness> under submissions/fe-corrective/ and fix all nine broken frontend apps (web/react/app-shell × easy/mid/hard). Use when asked to take the fe-corrective test, sit the frontend debugging benchmark, or generate a fe-corrective submission. |
| disable-model-invocation | true |
Test Taker (FE Corrective)
You are sitting the fe-corrective (frontend debugging) capability benchmark. Each question gives you a mostly-correct static web app with injected bugs. Fix them. Work independently and do your best.
Rules (read before starting)
- Read only from
questions/fe-corrective/<question>/. Never open, read, or reference reference-answers/ - it holds clean answers and checkers; consulting it invalidates the run.
- Write only inside your own submission folder. Do not modify
questions/, other submissions, or anything else in the repo.
- No build step, no CDN, no external network in the deliverable. Keep the same filenames.
- Budget roughly 6 minutes per question. Higher difficulty = more and subtler bugs.
Step 1 - Create your submission folder
Name it <llm>+<harness> in lowercase with no spaces. Examples: composer2.5+cursor, gpt-5+claude-code, sonnet+opencode.
submissions/fe-corrective/<llm>+<harness>/
q1-web-easy/ q1-web-mid/ q1-web-hard/
q2-react-easy/ q2-react-mid/ q2-react-hard/
q3-app-easy/ q3-app-mid/ q3-app-hard/
Step 2 - Answer each question
Read each PROMPT.md and the broken sources, then write the fixed files into the matching submission subfolder (same filenames).
| Question | Typical files to fix | What you are doing |
|---|
q1-web-* | index.html, app.js, (styles.css on easy) | vanilla product UI / checkout / kanban bugs |
q2-react-* | index.html, app.js (keep vendor/mini-react.js) | mini-react todo / cart / wizard bugs |
q3-app-* | index.html, app.js | browser app-shell (hash routes, tabs, offline notes) |
Constraints
- Static HTML/CSS/JS only. Entry file remains
index.html.
- For React tasks, use the provided local
vendor/mini-react.js — do not add CDN React.
- Apps must work when served over
http://localhost (Playwright grades them headlessly).
- Scoring is binary: 10 if the checker PASSes, else 0 — fix everything the prompt requires.
Step 3 - Finish
Verify all nine submission subfolders have their fixed sources, then report your submission folder name. Do not grade yourself.
