GenAI development techniques¶
A comparison of ways to organize and direct AI coding agents, for developers, tech leads, and engineering managers choosing an approach.
Start here¶
Choosing Your Approach — Match a technique to your team size, project type, industry, and development activity.
Overview and comparison matrix — How the approaches differ, with a comparison table and category summaries.
What this site covers¶
Compare workflows, reusable skills, and governance controls using dated research findings. Models and runtimes appear only where they help explain a technique. This is not a ranking or an adoption survey.
Detailed guides¶
See Research Methodology for sources, inclusion criteria, evidence labels, and update policy.
Spec-driven development¶
Star counts are snapshots observed on October 27, 2026. They are awareness signals, not adoption evidence.
| Technique | Description | Stars |
|---|---|---|
| GSD (Get Shit Done) | Spec-driven workflow continued as GSD Core after the original repository was archived | ~10.9K* |
| Spec Kit | GitHub's official toolkit for spec-driven development — specs → plans → tasks | ~143K |
| OpenSpec | Change-centric SDD with delta specs, broad multi-tool support, and a vendor-neutral skills target | ~73K |
* GSD migrated from an archived repository with ~64.7K stars to open-gsd/gsd-core; the successor repository started its own count.
Multi-agent orchestration¶
| Technique | Description | Stars |
|---|---|---|
| Squad | Coordinator-based multi-agent orchestration with persistent memory, casting, and ceremonies | ~3.3K |
| BMAD | AI-driven agile framework with specialized roles, structured phases, and modular workflows | ~55K |
Skill-based development¶
| Technique | Description | Stars |
|---|---|---|
| Superpowers | Mandatory skills methodology — TDD, subagent-driven development, and review enforcement | ~306K |
Autonomous iteration¶
| Technique | Description | Stars |
|---|---|---|
| Ralph | Autonomous bash-loop methodology — tests as backpressure, git as memory, tool-agnostic | Community |
Enterprise AI-native SDLC¶
| Technique | Description | Stars |
|---|---|---|
| HVE | Microsoft ISE's RPI workflow with constraint-based governance and a growing repository-level skills inventory | ~1.6K |
Cross-cutting¶
| Technique | Description |
|---|---|
| Context Engineering | The practice of structuring project context via rules files across an 8-layer model |
Skills and agent frameworks¶
These pages cover the infrastructure and libraries that support the techniques above. Superpowers is listed under skill-based development because it adds a mandatory workflow.
| Document | Description |
|---|---|
| State of the Skills Market | Market overview, taxonomy, and the shift toward versioned, managed skill packages |
| Skill Libraries | mattpocock/skills, addyosmani/agent-skills, anthropics/skills, vercel-labs/agent-skills, and watchlist collections |
| Orchestrating Modes | pstack/poteto-mode, canonical-source and port-drift caveats |
| Distribution and Governance | Agent Skills specification, marketplaces vs. vendored installs, provenance, permissions, and rollout guidance |
| Choosing a Skills Approach | Decision guide across libraries, orchestrating modes, methodologies, and native runtimes |
The Decision Guide provides a short selection table.
Building the site¶
Install requirements-docs.txt, run python scripts/build_docs.py, then mkdocs build --strict.
Diagram sources live in diagrams/ as standalone HTML files with inline SVG and CSS. A standalone Markdown link labeled Diagram: ... embeds that file's <figure> during staging; use diagrams/name.html from root pages or ../diagrams/name.html from technique pages. The original Markdown remains readable outside MkDocs. Figures follow the site's light/dark theme and scroll locally on narrow screens.
Run python -m unittest discover -s scripts -p 'test_*.py' to check diagram embedding.