Nnenna Ndukwe, Head of Developer Relations at Qodo AI, presents a comprehensive approach to building quality gates into agentic coding workflows, emphasizing structured AI-assisted software development. She introduces Qodo’s evolution from an AI code testing platform to a broader focus on code quality and management throughout the software development lifecycle (SDLC). Ndukwe highlights the importance of defining and codifying code quality standards that AI agents can consistently apply across various tools and stages, from planning and development to code review and deployment. She encourages developers to engage with a shared repository designed to demonstrate these principles in practice, using Codo AI as a platform for integrating rules, skills, and verification processes.
The workshop centers around a practical example: the Signal Pay payment API repository, which participants are invited to fork and clone. This repository includes detailed rules, skills, and templates to guide AI agents through planning, implementing, testing, and reviewing code changes with a focus on quality. Ndukwe stresses the importance of early and frequent quality checks using a combination of deterministic tools like linters and static analyzers alongside AI-driven reviews. The workflow includes local verification gates, continuous integration (CI) checks, and automated pull request (PR) reviews facilitated by Codo, ensuring that code adheres to predefined standards before merging.
A key part of the process is the use of structured planning templates that help define the scope, behavior scenarios, verification gateways, and failure recovery rules for each coding task. These templates enable AI agents to understand the expected behavior, relevant rules, and testing requirements before writing any code. Ndukwe demonstrates how agents can be instructed to generate high-level implementation plans and detailed build session execution plans, fostering consistency and clarity in AI-driven development. She also discusses the integration of various coding skills and rules, including those sourced from popular developers, to enhance code quality and maintainability.
During the code review phase, Codo AI automatically analyzes pull requests against organizational rules stored in a centralized portal, categorizing issues by severity and relevance. Ndukwe illustrates how this system can block merges for critical violations while allowing flexibility for less severe issues. The platform also supports automated fixes through agent skills, which can resolve identified problems either automatically or with human oversight. This iterative review and correction cycle helps maintain high code quality and reduces the risk of introducing bugs or security vulnerabilities.
In closing, Ndukwe encourages developers to adopt similar structured workflows and context management practices, whether using Codo or other tools, to improve AI-assisted coding outcomes. She highlights the value of combining deterministic verification tools with AI capabilities and recommends exploring open-source alternatives like Packmind for managing rules and skills across teams. The session concludes with a Q&A addressing practical concerns about skill effectiveness, local versus cloud deployment options, and the evolving capabilities of AI coding agents, underscoring the ongoing need for careful integration of AI into software engineering processes.
