Skip to content

Repository files navigation

Agentic Engineering

Agentic Engineering — Patterns, Decisions and Compositions

Patterns, decisions and minimal examples for governed, spec-driven, multi-harness agentic software engineering.

Purpose

This repository captures what we learned while building software with AI coding agents under human supervision. It is a technical knowledge base, not a tutorial or a product.

The organizing model is:

ToolDecisionPatternCompositionExample

  • Tools document external building blocks and their relevant properties.
  • Decisions record architectural choices using the MADR format, with context, alternatives, and consequences.
  • Patterns describe reusable approaches: how they work and when to apply them.
  • Compositions show how multiple patterns work together to form a coherent workflow.
  • Examples are minimal, neutral artifacts that demonstrate a concept without requiring project-specific context.

Repository Map

agentic-engineering/
├── tools/               # Tool assessments (OpenSpec, Superpowers, OpenCode, Claude Code, Codex)
├── decisions/           # MADR-style architectural decision records
├── patterns/            # Reusable patterns for multi-agent engineering
├── compositions/        # How patterns combine into workflows
├── examples/            # Minimal neutral examples
├── LICENSE
└── README.md

What This Is Not

  • Not a toolkit. Installable configurations, scripts, and harness adapters live in agentic-dev-toolkit.
  • Not SpecRivet documentation. When the SpecRivet product repository is published, it will reference patterns and decisions from here, but it will be independently understandable, installable, and usable. There is no runtime dependency from SpecRivet to this repository.
  • Not a personal diary or chronological journey. Documents are organized by concept, not by date of discovery.
  • Not a dump of research notes. Every document was rewritten from scratch to extract only the generic engineering principle.

Relationship to agentic-dev-toolkit

This repository agentic-dev-toolkit
What have I learned/designed? What can I reuse/install?
Conceptual: patterns, decisions, examples Operational: scripts, configs, adapters
Docs-first, minimal executable content Executable-first, documentation supports usage

The two repositories are complementary. This one explains why a certain approach works; agentic-dev-toolkit provides the assets that implement it.

Philosophy

  • Docs-first. Code and configuration exist only when they help prove or explain a pattern.
  • Rewrite, don't copy. Every document was rewritten to extract the generic principle from its original context.
  • Evidence-aware. Documents distinguish observed behavior, design rationale, and proposed reusable patterns.
  • Minimal examples. Examples are small, neutral, and understandable without project-specific context.

License

MIT

About

Patterns, decisions and minimal examples for governed, spec-driven, multi-harness agentic software engineering.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors