On 31 August 2026, our Technical Director and founder Nick Allen presented at the Sitecore User Group Jaipur, an online meetup for Sitecore practitioners.
The session, Sitecore development with specs, skills and agent memory, looked at how to get consistent, reviewable output from AI coding agents on real Sitecore projects. It was built around a measured experiment: one Sitecore ticket, run through Claude Code nine times, first on its own, then with a written specification, then with house skills added to the repository.
The talk covered:
- Specs: how a specification closes the questions a business ticket leaves open, and why the agents stopped asking for clarification once one was in place.
- Skills: how house conventions that never appeared in the prompt were applied in every run, and why an installed skill can still be ignored.
- Memory: what agents should carry between projects, and why memory needs a review discipline of its own.
The session is available to watch in full, including the Q&A:
Watch the recording on YouTube →
For the findings in written form, with the results from each stage and practical recommendations for Sitecore teams, read Specs, skills and memory: getting consistent Sitecore code from AI agents.
Thank you to the Sitecore User Group Jaipur for hosting us, and to everyone who joined and asked questions.
