Regression Log for Prompt Changes
Track every prompt change and its effect on test results.
Added October 15, 2026. Written with AI assistance and checked for clarity and placeholders; not reviewed by a lawyer, accountant, clinician or other professional. How these prompts are made
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Create a regression log process for prompt changes to [AGENT_OR_ASSISTANT]. The test suite I have: [TEST_SUITE]. Who changes prompts: [EDITORS]. Release process: [RELEASE_PROCESS].
Design the log with fields: change id, date, author, the exact prompt diff (before and after), the reason, the hypothesis ('this will reduce X'), tests run, results by category, regressions found, decision (ship, revise, revert), and follow-ups. Write the procedure: change one thing at a time, run the full suite, compare to the baseline, review any case that moved from pass to fail, and require a second person for changes to safety rules. Provide a comparison table layout, rules for noisy tests (run three times), and a weekly summary format. Add an example of three entries for an invented assistant, including one reverted change.
Explain how to version the prompt files so a past release can be reproduced.
Add a rule for noisy results: run each test three times, record the pass count out of three, and count a case as failed only if it fails twice. Include an example comparison table with baseline 36 of 40 passed and new version 38 of 40 passed, with the two newly failing cases listed for review.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.