Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned# Axiom ReleaseGuard
AI-Assisted Release Readiness for the Life After Code Era
Inspiration
Building software is only part of the development journey. Before releasing an application, developers also need to verify that essential project files exist, automated checks pass, and outstanding risks are understood.
As an AI-assisted developer and founder of Tierra AI Labs, I wanted to explore how GitLab Duo Agent Platform could help developers move from writing code to making better-informed release decisions.
That idea became Axiom ReleaseGuard: a lightweight release-readiness assistant combining a TypeScript application, GitLab CI/CD, and a custom AI agent.
What it does
Axiom ReleaseGuard provides three complementary capabilities:
1. Basic file-readiness analysis
The application checks whether five foundational repository files are present:
README.mdLICENSEpackage.json.gitlab-ci.yml.gitignore
It displays the results in a local web dashboard and provides a basic file-readiness score. The score measures file presence only, not code quality or security.
2. Automated CI/CD validation
A GitLab CI/CD pipeline runs two jobs:
typescript-build— compiles the TypeScript application.automated-tests— runs automated application tests.
This gives developers verifiable build and test results within GitLab.
3. AI-assisted release review
I created a custom Axiom ReleaseGuard agent using GitLab Duo Agent Platform.
The agent can inspect repository files and pipeline information, summarize verified findings, identify missing evidence and release risks, and recommend next steps.
Rather than treating a passing pipeline as proof of production readiness, the agent distinguishes successful checks from issues that still require investigation.
How I built it
I developed the project on Windows using PowerShell and AI-assisted development workflows.
The application uses TypeScript, Node.js, and tsx, with a lightweight local HTTP dashboard. I wrote automated tests using the Node.js test runner and configured a GitLab CI/CD pipeline using .gitlab-ci.yml.
I first developed the application in the Tierra AI Labs GitLab repository, then integrated it into the official GitLab Transcend hackathon Showcase project through a feature branch and merge request.
After merging the application into main, GitLab pipeline #2946925608 completed successfully, with both the TypeScript build and automated test jobs passing.
I also configured a custom GitLab Duo agent with repository and pipeline inspection tools. I tested the agent against the Showcase project and asked it to verify the required files, inspect the successful pipeline, identify risks, and recommend next actions without modifying any files.
The agent confirmed the five files were present and both pipeline jobs had passed. It also identified limitations, including the absence of demonstrated security scanning and deployment verification.
Challenges I faced
Integrating with the official hackathon repository
My application began in a separate Tierra AI Labs repository. I needed to move the working implementation into the official Showcase project while preserving its existing history and using the correct merge request workflow.
Configuring GitLab Duo
Access to GitLab Duo differed between projects. I worked through those limitations and successfully tested the custom agent in the official hackathon workspace.
Balancing automation with accuracy
One important challenge was ensuring the project did not confuse successful file checks or CI jobs with a complete production-readiness assessment. I designed the agent instructions and project documentation to separate verified evidence from unverified assumptions.
Working within a $0 budget
I focused on tools and workflows available without purchasing additional subscriptions or credits.
What I learned
This project strengthened my understanding of GitLab repositories, branching, merge requests, CI/CD pipelines, automated testing, and GitLab Duo Agent Platform.
More importantly, I learned that an effective AI development assistant should not simply announce that everything is ready. It should explain what was checked, what passed, what remains unknown, and what should happen next.
Accomplishments
- Built a functioning TypeScript release-readiness application.
- Created a local dashboard showing five file-presence checks.
- Implemented automated tests.
- Successfully ran GitLab CI/CD build and test jobs.
- Merged the application into the official hackathon Showcase repository.
- Created and tested a custom GitLab Duo release-review agent.
- Demonstrated evidence-based AI recommendations without modifying repository files.
- Produced a silent project demonstration video.
What's next
Future development could include dependency and security scanning, broader test coverage, deployment verification, and automatically triggered GitLab Duo workflows.
The current prototype demonstrates AI-assisted release review, not a fully autonomous release pipeline or a certification that software is production-ready.
Why it matters
Axiom ReleaseGuard helps developers understand the difference between code that passes basic checks and a release that has enough evidence to justify moving forward.
Built by Tierra AI Labs for the GitLab Transcend October 2026 — Life After Code Hackathon.
What's next for Axiom ReleaseGuard
Built With
- agent
- ci/cd
- css
- duo
- git
- gitlab
- html
- javascript
- node.js
- platform
- powershell
- tsx
- typescript
Log in or sign up for Devpost to join the conversation.