How to Build Quality Gates into Agentic Coding Workflows — Nnenna Ndukwe, Qodo AI
Summary
This workshop details how to implement structured quality gates into agentic coding workflows, moving beyond simple local testing. Using a sample FastAPI payments application (SignalPay), the speaker demonstrates a comprehensive process that integrates planning, local verification, and automated code review (PR review) using tools like Qodo/Codto. The core principle is operationalizing context—codifying standards, rules, and skills—to ensure AI agents adhere to defined quality and governance standards across the entire Software Development Life Cycle (SDLC).
Key takeaways
-
Structured AI Coding Workflow
6:49
A robust workflow requires defining and codifying code quality standards (rules) and expected behavior (skills) in a central context management layer, rather than relying solely on individual repository knowledge.
-
The Quality Gate Process
11:51
The process involves multiple gates: 1) Planning (creating high-level and build session execution plans); 2) Local Verification (using deterministic tools like linters, Pytest, Bandit); 3) PR Review (Codto/Qodo analysis); and 4) Remediation (using agent skills to automatically resolve identified issues).
-
Operationalizing Context
8:26
The system must ingest rules and skills from a central source (context management layer) to apply them consistently during planning, code generation, and code review, ensuring the agent is aware of all relevant constraints.
Technical details
-
Workflow Setup & Tools
600s
The demonstration uses a SignalPay FastAPI payments API. Prerequisites include Git, GitHub CLI, Python, UV, Node.js/npm, and a coding agent (e.g., Codeex, Claude). The workflow utilizes Qodo/Codto for centralized rule and skill ingestion.
-
Planning and Execution
1200s
The planning phase requires generating both a high-level plan and a detailed build session execution plan. These templates must specify the expected behavior, applicable rules, required skills, and the 'Definition of Done' (DoD) to guide the agent.
-
Local Verification Gates
1200s
Local checks must be deterministic and include tools like linters, Pytest, Bandit, and static analysis (e.g., using `grep`) to ensure local contracts are preserved before PR submission.
-
Code Review and Remediation
1800s
When a PR is opened, the platform (Codto) analyzes the code against ingested rules (e.g., 'Python functions must all have a doc string attached to it'). Issues can be automatically resolved using a dedicated agent skill (PR resolver), completing the code review loop.
Mentioned resources
Channel & topics
Watch on YouTube · Back to latest
This independent, AI-assisted summary is provided for commentary and informational purposes. It may contain errors or omit important context. Please watch the original video for the creator's complete presentation. Video, thumbnail, and related copyrights belong to their respective owners.