A complete, defensible history for your Knowledge Agent's prompts and skills
Every change to a Knowledge Agent's system prompt and skills is now versioned.
Previously, saving a change overwrote what was there, with no record of how the agent was configured before. Now every save creates a new version. A History button above the prompt in agent settings, and in each skill's options menu, opens a table of every version showing who made the change and when. From there you can view any version, compare two side by side, or revert to a previous version and add change notes. History covers changes made since May 2026.If you own or manage Knowledge Agents, this is for you
. It's especially valuable for teams with audit, compliance, or regulatory obligations that need to show exactly how their AI was configured at any point in time. It's just as useful as a safety net if you iterate on prompts often. Access follows the same permission as editing the agent's settings, and the people chatting with your agent won't see anything change.The history is built to be defensible. Every version is immutable and captures the full state of the prompt or skill, not just the text that changed
. A revert is recorded as its own event, so the history is never rewritten. Answer details now also show the exact prompt and skill versions used to generate each answer. That means you can trace a questioned answer back to the configuration that produced it, even after a dozen changes since.Open a Knowledge Agent's settings and click History above the prompt
. Start small. Compare the current version of a skill you've been iterating on against an earlier one, or open the details of a recent answer to see which version produced it. From there, the history builds itself every time you save.
