video1.0<iframe src="https://www.loom.com/embed/0ed6398636b94afaa225ede84627da8f" frameborder="0" width="2560" height="1920" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>19202560Loomhttps://www.loom.com19202560https://cdn.loom.com/sessions/thumbnails/0ed6398636b94afaa225ede84627da8f-52a65bcc5d4823b0.gif378.896This Loom demonstrates how to use Tali Stash to offboard prompt development and prompt versioning between Claude with Clay. The speaker develops prompt versions in Cloud against Kernel’s customer corpus and then pushes the revised prompts into Stash via a Tali utility endpoint that can include a self-destruct timer, keeping the same external IDs across versions. In Clay, they retrieve the latest prompt from Stash by stash ID in a custom table and feed the system prompt, user prompt, and output schema into research columns on a schedule (at most once a day). They also note Stash acts like a GitHub or Redis for prompts by preserving prior versions for client work and easy reversion.