Vibe-code a pull request dashboard across Azure DevOps repos
Author used ChatGPT for architecture and the Junie agent in WebStorm to build a dashboard aggregating PRs across four repos, flagging approvals, conflicts and comments, in about 3 hours. Task: The author defined dashboard goals for ChatGPT 5 to generate an architecture, then fed this into the Junie agent in WebStorm to create project guidelines and code. Development proceeded step-by-step: layout, dummy data, Azure API calls, YouTrack integration, and features. The author reviewed generated code and prompted for refinements without manual editing, ensuring PATs remained local.; tools: ChatGPT. Human role: Not stated in the source; limitations: The workflow relies on AI agents that may struggle with large existing codebases or produce inconsistent patterns. Success required iterative prompting to enforce code reuse and centralized configuration. Security depends on keeping Personal Access Tokens (PATs) local and out of the codebase. The author's prior experience likely aided project management, though they did not write code..
Tools
Published on Agent Work Atlas:
Overview
Author used ChatGPT for architecture and the Junie agent in WebStorm to build a dashboard aggregating PRs across four repos, flagging approvals, conflicts and comments, in about 3 hours.
Workflow
The author defined dashboard goals for ChatGPT 5 to generate an architecture, then fed this into the Junie agent in WebStorm to create project guidelines and code. Development proceeded step-by-step: layout, dummy data, Azure API calls, YouTrack integration, and features. The author reviewed generated code and prompted for refinements without manual editing, ensuring PATs remained local.
Reported result
A functional dashboard with three views was built in approximately 3 hours, aggregating PRs across four repos. The team adopted it daily after an additional hour of feature additions. Most steps succeeded on the first try, though the author noted Junie occasionally produced repetitive or inconsistent code that required prompt adjustments.
Human role
Not stated in the source
Limitations
The workflow relies on AI agents that may struggle with large existing codebases or produce inconsistent patterns. Success required iterative prompting to enforce code reuse and centralized configuration. Security depends on keeping Personal Access Tokens (PATs) local and out of the codebase. The author's prior experience likely aided project management, though they did not write code.
Try it
Act as a developer building a PR dashboard. First, describe the requirements to ChatGPT: aggregate active PRs from [replaceable: list of Azure DevOps repos] into a single view, linking to YouTrack and flagging approvals/conflicts. Use PATs stored locally. Second, input the architecture into the Junie agent in WebStorm to generate a Readme and project structure. Third, instruct Junie to implement step-by-step: 1) Basic layout with dummy data, 2) Azure DevOps API integration, 3) YouTrack linking, 4) Feature flags for attention items. After each step, review the code for consistency and prompt Junie to 'maximize code reuse and centralize configuration' if repetition is found. Do not manually edit code; use prompts to fix issues. Verify PATs are not in the repo. Measure success by time to first working view and percentage of steps requiring correction. Start with a small sample of 2 repos for a dry run before expanding.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.