💡 How to use these prompts:
Replace everything in [BRACKETS] with your specific details before sending.
Click Copy to copy any prompt to your clipboard instantly.
Search this repository for every place a model name is specified: editor and Copilot configuration, agent definitions, CI and workflow files, environment variables, infrastructure as code, and any hardcoded string in application code. Output a table of file path, line, the model named, and whether that reference would error or silently fall back if the model disappeared. Sort so the silent-fallback cases come first.
I am migrating off [OLD MODEL] to [NEW MODEL] by [DATE]. Produce a migration table listing every call site from the audit, the change required at each, whether the change is mechanical or needs judgement, and a rough effort estimate. Flag any call site where the prompt was tuned to the old model and may need retuning rather than a find and replace.
From this codebase, identify the five tasks where model quality matters most and write them up as a regression set I can run against both the old and the new model before the cutoff. For each task give the input, what a good output looks like, and what specific failure I should watch for. Prefer tasks that are slow and painful over tasks that are easy to score.
Write a check I can run in CI that fails loudly if any model reference in this repository resolves to a model other than the one explicitly configured. It should compare the model actually used in a response against the model requested, and fail the build on mismatch rather than logging a warning. Explain where in my pipeline this check should sit.