The challenge
Standards that lived in people’s heads.
Every developer had their own way of building blocks, naming things and checking work before launch. Reviews caught some of it, but consistency depended on who built the site and how busy they were.
AI assistants could speed the work up, but only if they knew the team’s conventions. Left alone, they produced code that looked fine and broke house rules.
The approach
How I tackled it.
-
Write the rulebook down
Turned the team’s coding and build standards into documents an AI could follow, pulled fresh from the standards repo.
-
Package skills
Built repeatable skills for common jobs: building a block, assembling a page, entering content, running a launch checklist.
-
Add review agents
Added read-only agents that review code for standards, security issues and QA problems and report findings for a person to triage.
What I built
Shared skills and agents for the whole team.
- Build skills Guided workflows for blocks, pages, theme setup and content migration that follow house rules by default.
- Standards reviewer Checks new code against the current coding standards before it merges.
- Security reviewer Looks for unescaped output, missing permission checks and unsafe input handling.
- QA crawler Crawls dev sites for broken links, console errors, missing alt text and layout issues before client review.
The results
What changed.
Builds start from the same playbook, reviews catch the same issues every time, and new developers get the team’s standards from day one.
- Best practices built in Standards are applied while the work is done, not just at review.
- Consistent builds across projects Every site follows the same structure and conventions.
- Automated standards reviews Agents flag issues for a person to triage before merge or launch.
Tech stack
- AI agents
- Skills
- Git
- WordPress
- PHP
- Playwright
- Lighthouse
