Agent-ready engineering standards at scale

Business problem

As Baker Hill scaled Angular and C# delivery, with engineers adopting AI-assisted coding along the way, teams felt the same friction from several directions at once: review churn, inconsistent patterns across repos, slow onboarding to “how we build here,” and rework when agent-generated changes did not match architecture or product conventions.

Role

Gunnar set the decision framework for what belongs in shared standards and how teams extend them: Cursor rules and skills, Angular schematics, review expectations, and CI hooks. He defined what to capture and how to add it, coached reviewers on using the material, reviewed team contributions, and filled gaps himself.

He evangelized the approach when people brought problems that standards or agents could solve; adoption took time before “encode it in rules” became default thinking.

Stack & approach

Cursor rules and skills in repos; Angular schematics for repeatable structure; review guidance aligned with architecture standards; CI hooks that reinforce the same bar. Content encodes frameworks, branching and versioning practice, infrastructure constraints, and internal vocabulary so agents interpret company-specific shorthand, not only generic best practices.

Loop

Tradeoffs

Standards must be prescriptive enough to improve AI-assisted output without freezing team autonomy. Gunnar’s model: a clear framework for what gets documented, peer review of additions, and ongoing updates as models and tools change, rather than one static style guide nobody maintains.

Outcome

Stronger pattern consistency and higher-quality AI-assisted pull requests, with less review churn before code reaches “good enough.” Teams reach acceptable quality faster, which also cuts wasted agent iteration and token use. The same corpus laid groundwork for an agentic SDLC pipeline that can reason about Baker Hill’s frameworks, processes, and environments, not just isolated snippets.