| name | test-taker |
| description | Take the LLM-agent capability benchmark - create a submission folder named <llm>+<harness> and answer all three questions (offline dashboard, CSV analysis, architecture design). Use when asked to take the test, sit the benchmark, or generate a benchmark submission. |
| disable-model-invocation | true |
Test Taker
You are sitting a timed capability benchmark. Produce your own answers to all three questions and save them as files. Work independently and do your best.
Rules (read before starting)
- Read only from
questions/. Never open, read, or reference reference-answers/ - it is the answer key, and consulting it invalidates the run.
- Write only inside your own submission folder (see Step 1). Do not modify
questions/, other submissions, or anything else in the repo.
- Budget roughly 10 minutes per question. High complexity is expected.
Step 1 - Create your submission folder
Name it <llm>+<harness> in lowercase with no spaces - your model name joined to the harness you run in. Examples: composer2.5+cursor, gpt-5+claude-code, sonnet+opencode. If you cannot determine your model or harness, ask the operator once, then proceed.
Create these folders:
submissions/<llm>+<harness>/
q1-offline-dashboard/
q2-csv-analysis/
q3-architecture-design/
Step 2 - Answer each question
Read each PROMPT.md, then write the exact deliverable into the matching subfolder of your submission:
| Question | Read | Produce |
|---|
| Q1 | questions/q1-offline-dashboard/PROMPT.md | dashboard.html |
| Q2 | questions/q2-csv-analysis/PROMPT.md and sales_data.csv | analysis.md |
| Q3 | questions/q3-architecture-design/PROMPT.md | architecture.md |
Follow every constraint in the prompt exactly - for example, Q1 must be a single file that works offline with no external/CDN resources. Do not add extra files unless a prompt asks for them.
Step 3 - Finish
Verify all three deliverables exist at the correct paths, then report your submission folder name. Do not grade yourself.
