Agentic Engineering Starter Pack

This repository is a small documentation starter pack for software projects developed with autonomous coding agents.

The useful thing is the agent operating context under starter/. Copy that context into a project, make it local, and let it become the shared operating model for humans and coding agents working in that repository.

The docs should preserve not only what the project does, but what kind of product it is trying to become, so future agents do not collapse an ambitious product direction into a smaller familiar implementation pattern.

Core Idea

Coding agents work best when the repository gives them a small amount of stable context:

• what the product is trying to become
• how the project works today
• how agents should interpret goals and make decisions
• what durable decisions, caveats, and lessons future agents should inherit

The starter keeps those ideas in four copied surfaces:

AGENTS.md              agent operating guide
docs/product-intent.md product north star
docs/approach.md       current project approach
docs/records/          durable project memory

This repository can also carry optional skills under skills/. Skills are not part of the default copied project context. Use them when a project needs a specialized agent behavior beyond the general operating guide.

How To Use It

Copy only the agent operating context into a new or existing project:

starter/AGENTS.md -> AGENTS.md
starter/docs/     -> docs/

Do not copy this repository's README into the project. A project README should be written for the actual project: what it is, how to run it, how to use it, and how to operate it.

After copying the starter context, replace the placeholders with the project's real context. The docs should start simple and become more specific as the product, architecture, and working model become clearer.

The copied docs are meant to be read and updated by coding agents as they work. They should help agents act with initiative while staying grounded in the project's actual goals, code, decisions, and constraints.

Optional Skills

Use skills/ for compact skill files that teach agents a specialized engineering posture.

Current skills:

skills/build-agentic-systems/
skills/product-ideation/

build-agentic-systems is for projects that build LLM agents, tool-using agents, subagents, agent runtimes, workflow agents, autonomous coding agents, or systems where models choose actions. It communicates the core philosophy:

The LLM owns strategy.
The runtime owns semantics.

product-ideation is for early product, feature, company, workflow, or system shaping before implementation. It keeps agents focused on the customer, pain, buyer, user, promise, adoption path, trust, value, and key assumptions before naming solution machinery.

Keep skills small. Start with a single SKILL.md; add references, scripts, or assets only when repeated use proves that the extra structure helps.

What To Keep

Keep the starter small and general.

Prefer improving the copied operating context over adding new documentation categories. Add more structure only when it helps future agents understand the project and make better decisions.

The goal is not to create a process manual. The goal is to give each project enough shared context that coding agents can build, verify, document, and deliver work without relying on hidden chat history.