ADSM / GitHub Flows Pilot
A bounded pilot for one repository, one workflow, one task type, and one reviewable result. The goal is to test whether the project can use controlled AI-agent development in practice.
ADSM / GitHub Flows
Controlled AI-agent development for web applications
I build GitHub-based workflows where product changes move from issue to agent execution, pull request, validation, and publication under documented control.
Public proof: GitHub Issue → AI agent → published page
On TeqFW Demo Pages, any GitHub user can open a constrained issue requesting a demo page. If the request fits the documented boundaries, GitHub Flows starts an agent, the agent creates a bounded page, and the result becomes a published artifact.
This is not just a portfolio item. It is a public proof that controlled product evolution can run through GitHub issues, documented constraints, agent execution, review, and publication.
What this proves
- User intent can enter a product through GitHub Issues.
- AI agents can work inside documented constraints.
- GitHub provides traceability through issues, labels, comments, pull requests, and validation evidence.
- The product owner keeps control over allowed changes, review points, and publication.
What I offer
Agent-ready repository setup
Repository structure, context documentation, AGENTS.md, GitHub labels, workflow profiles,
execution environment, and basic rules for agent work.
Controlled web application development
Small web applications, MVPs, PWAs, internal tools, and self-hosted products developed through documented context, GitHub workflow, and agent-assisted iterations.
Early pilot projects available
I am currently accepting a limited number of early pilot projects. The goal is to apply ADSM and GitHub Flows to real repositories through small, bounded, reviewable engagements.
Start small. Prove the workflow. Improve it iteratively.
Method stack
ADSM is the method.
Agent Driven Software Management defines how context, documentation, constraints, validation, and evidence keep agent-driven development under control.
GitHub Flows is the execution layer.
It connects GitHub events, issues, labels, pull requests, server-side agent execution, comments, and validation into a controlled workflow.
Tequila Framework is the application platform.
It provides a JavaScript foundation for building agent-friendly web applications: DI, modular structure, web server, configuration, logging, persistence layer, and related components.
Selected proof projects
TeqFW Demo Pages
Proof value: public issue-driven agent workflow with published results.
GitHub Flows / GitHub Flows App
Proof value: orchestration layer for agent execution through GitHub infrastructure.
Tequila Framework
Proof value: JavaScript platform for agent-friendly applications.
This website / TeqCMS
Proof value: sites and content systems developed through git, documentation, and AI-agent workflows.
The book behind the method
My Leanpub book Agent-Driven Development explains ADSM: how to stay in control while delegating development work to AI agents.
About Alex
I am Alex Gusev, an independent web developer from Latvia with 25+ years of experience in web application development. My current focus is controlled AI-agent development, ADSM, GitHub Flows, and JavaScript application architecture.
Want to test controlled AI-agent development on your repository?
The best first step is a bounded pilot: one repository, one workflow, one real task, one reviewable result.