What it does: Writes 5 test cases from your real work and a score sheet, so you can run 2 models side by side in Hugging Face's Inference Playground.
Use it when: You're picking an AI model for a job and every review says something different.
You get: 5 test inputs from your work, what a good answer looks like for each, a score sheet, and the winner once you paste the answers back.
You are my test planner, a pro who picks AI models by testing them on real work, not on reviews. Goal: write 5 test cases from my real work and a score sheet, so I can run 2 models side by side in Hugging Face's Inference Playground and pick the winner on my own numbers. Context: I run a small business and need one model for one job. Reviews disagree because they test other people's work, so the only test that counts is one built from mine. First, check that you can read the 5 real examples I gave you for this job. Tell me how many you found. If they're missing, stop and ask me to paste them instead. Ask me these, then run. Skip any I already answered. 1. The job the model must do. (Default: guess it from my examples and tell me what you guessed.) 2. What must a good answer have? (Default: the right answer, and one line why.) 3. Which 2 models? Give the exact names from the Playground list. (Default: write [check: model A] and [check: model B], I'll pick them in the Playground.) Do this: 1. Check my examples for private details: names, emails, phone numbers, addresses. Swap any you find for fake ones and tell me how many you swapped. 2. Turn my examples into 5 test inputs: one easy, three normal, one hard. If I gave fewer than 5, build the rest from my examples and mark each one (built from example N). 3. For each test, write what a good answer looks like, using only what a good answer must have. 4. Write one system message to use for both models, under 60 words. 5. Make a score sheet: test number, model A score 0 to 2, model B score 0 to 2, notes. 0 = wrong, 1 = close, 2 = right. 6. Tell me how to run it in the Playground in 3 steps: same system message and same settings for both models, paste each test, copy each answer back here. 7. When I paste the answers back, score each one on the sheet only, add up both totals, and say which model wins and where each one failed. A tie goes to the model with fewer 0 scores. Give me the 5 tests with their good answers, the system message, the score sheet as a table, and the 3 run steps. Keep each good answer under 30 words. Example score row, match this shape: | 3 | 2 | 1 | Model B gave the right kind but skipped the one line why | Before you show me, check: - There are exactly 5 tests: 1 easy, 3 normal, 1 hard. - No real name, email, phone number or address is left in any test. - When you score, both totals add up from the rows. Rules: - Same system message and same settings for both models. - Score only on what a good answer must have. - Drafts only. Ask me before you share, post or change anything. - If an example has no right answer from me, write your best guess and mark it (guess). If an example isn't this job, skip it and tell me. - Use only what is in my data and what I tell you. Never make up a name, number, date, price, link or promise. - If something you need is missing, write [check: what is missing] in that spot and keep going. Last, save these steps as a skill called 'Test Two Models' so you do it this way each time I pick a model for a job. To run it again, just say: "Run my Test Two Models skill."
You need
Hugging Face![]()
Any AITested on sample data
Built on Fanatical Prospecting's rule to know your numbers.
Copilot version
There is no separate Copilot version of this one yet. Paste the prompt above into Microsoft 365 Copilot Chat.
See a real example
Tested on sample data Oct 29 2026, after the rewrite: a Claude agent ran it on made-up sample input (tool results simulated). A separate Claude agent judged the run and the card.
Watch: no video for this one yet.
AuthorCorey Leon