Playbooks

Turn a vague bug report into a failing test, and refuse to let the model touch the source until that test exists.

Define what the first agent must pass to the second, and prove nothing needed was dropped.

Convert a real incident into a replayable test that fails if the same bug returns, including after a model update.

Add termination conditions to a retrying agent so it fails loudly at a known point rather than burning budget quietly.

Turn a pile of logged bad outputs into a scored eval set, so the next prompt change is measured instead of guessed.

Produce a rubric with criteria, bands, and descriptors that actually discriminate between the samples you already have.

Turn a clinical discharge summary into a plain language version that keeps every instruction, dose, and warning intact.

Take a refusal notice and produce a table of every reason cited, what would answer it, and which drawing needs to change.

Turn 100 GitHub commits into IP assignment log showing who created what and when for cap table.

Feeds a lengthy PDF report into an AI model and gets back a clean, skimmable weekly digest.

Confirm the schema before migrating.