Kyle Penfound is a solutions engineer at Dagger with more than a decade of experience in DevOps, platform engineering, and release engineering. He builds agentic continuous integration systems that give coding agents the same containerized environments, project-specific tests, and review requirements used to deliver production software.
In 2023, Penfound advocated writing CI pipelines as portable, testable code, replacing brittle, platform-specific YAML with reproducible workflows that run consistently on laptops and CI infrastructure. He subsequently extended that approach to AI agents: existing functions for building applications, running tests, and enforcing engineering standards become tools an autonomous contributor can use.
His greetings-api repository demonstrates agents that implement feature requests, open pull requests, incorporate review feedback, review proposed changes, and diagnose failing tests. Dagger modules and GitHub Actions connect these capabilities to the application’s existing delivery process.
Agent tool scoping: Agents need enough flexibility to solve unfamiliar problems without unrestricted permissions or an overwhelming tool inventory. If a project already specifies how tests run, an agent should receive that dedicated testing function instead of inventing shell commands.
Independent verification of generated code: Asking a model to run tests is insufficient; the surrounding workflow must execute those checks separately before accepting its changes. Container isolation protects local files, while execution traces reveal tool calls, failures, and opportunities to improve prompts or tool descriptions.
Automated contributions, accountable review: In his writing on AI-era continuous integration, Penfound argues that coding agents increase pressure on maintainers by accelerating contributions faster than qualified human review. Portable checks, documentation, linting, meaningful tests, and automated feedback can improve incoming changes without eliminating human responsibility.
At AI Engineer World’s Fair 2025, Penfound and fellow Dagger ecosystem engineer Jeremy Adams-Casañas demonstrated a GitHub issue-to-pull-request workflow in which a scoped agent inspects a repository, modifies code, runs existing tests, and submits changes for review.
▶ Watch ↗