AI Coding Agent Development Playbook

A practical toolkit for AI coding-agent workflows: a reusable development playbook, sample skills, sample agent instructions, and adaptation guides for making repositories easier for humans and agents to understand, coordinate, resume, and safely maintain.

Contents

• ai-agent-development-workflow-playbook.md - the reusable workflow playbook.
• guides/ - guides for adapting the playbook and sharing skills across coding agents, including optional project module mapping.
• samples/agent-instructions/ - root and module-level AGENTS.md samples.
• samples/module-maps/ - a project-neutral map registry and semantic module map example.
• samples/skills/ - sample skills split by domain, including general coding, documentation, Unity, Obsidian, and robotics categories.

How To Use This With An Agent

Attach ai-agent-development-workflow-playbook.md or point your agent at this whole repository, then use a prompt like this:

Please read the attached AI Coding Agent Development Workflow Playbook and this
sample repository. Then adapt the workflow to my project.

First inspect my repository. Identify the project type, language/toolchain,
frameworks, build/test commands, protected/generated/vendor folders, existing
docs, and any collaboration risks.

Then propose and implement a project-specific agent workflow:
- create or update AGENTS.md as the canonical standing-rules file
- create docs/current-status.md, docs/command-guide.md, and
  docs/architecture-baseline.md, or map those roles to existing project docs
- create docs/active-work.md, or an equivalent ownership tracker, if multiple
  people or agents may work at once
- decide whether recurring repository discovery justifies project module maps;
  if so, create a coverage registry and map modules independently
- require agents to load only the smallest applicable map, maintain affected
  relationships in the same change, and reconcile destination-branch
  instructions and maps during integration
- recommend project skills only for recurring tasks
- if useful, add one canonical skill and thin wrappers for Codex, Claude, and
  GitHub Copilot, following guides/sharing-skills-across-coding-agents.md

Treat the playbook as a set of roles and patterns, not a rigid folder template.
Rename, combine, or skip pieces when my project already has an equivalent place
for that information. Explain any adaptation you make.

If this is a Unity project, use the Unity samples only as patterns. Do not copy
any sample unless the project actually uses that architecture. Keep changes
scoped, preserve existing conventions, and tell me what you changed and what
still needs review.

For a first pass, ask the agent to draft the docs and skill recommendations before making sweeping changes. Once the structure looks right, have it create or update the files directly.

License

MIT. See LICENSE. The playbook, guides, and samples are intended to be copied and adapted for project-specific agent workflows.