AI-native development is not only a matter of asking a model to write code. Real project work also needs context, sequencing, validation, handoffs and decisions about when a person must review the result. Development systems make those operating rules visible so that AI assistance can fit into a repeatable workflow.
The ABVX catalogue groups these experiments into three related areas: workflow and orchestration, development surfaces and interfaces, and protocols and decision systems. SET, the AGENTS.md Generator and ABVX Agent Skills address operating structure and reusable procedures. ABVX Lab and AsciiTheme represent developer-facing surfaces. ID and Decision Map show how protocols can structure references and choices.
From prompts to operations
A prompt is one moment in a larger chain of work. An agent may need to inspect a repository, find the applicable instructions, choose a bounded method, change files, run checks and report what the evidence proves. When these steps are explicit, a team can review how work moved from request to result.
This does not mean automating every judgment. Some decisions depend on project owners, external approval or information a tool cannot verify. A well-designed workflow makes those boundaries clear instead of hiding them behind a successful code-generation step.
A connected development stack
Tools, interfaces and protocols serve different roles. Orchestration coordinates tasks; instruction files preserve project context; reusable skills package procedures; a development surface helps people inspect available tools; protocols give the system stable terms for identity, decisions or communication.
The projects here are practical work and ongoing experiments, not a claim that one stack fits every team. Explore the three groups below to see how each layer contributes to structured, inspectable AI-assisted development.