# Ground Action Composition

> Action composition. Reuse without deep chains.

- Skill: `attac-t/ground-action-composition` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add attac-t/ground-action-composition`
- Raw SKILL.md: https://api.skillmd.com/api/skills/attac-t/ground-action-composition/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: attac-t (https://skillmd.com/u/attac-t)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/attac-t/ground-action-composition

---


# Skill: Action Composition

> "Actions reduce the cognitive load that's introduced by such a system."

## The Standard

- **Flat Over Deep**: Maximum 3-4 levels of action calls.
- **Inject Dependencies**: Actions receive other actions via constructor.
- **Single Responsibility**: If an action does too much, split it.
- **Copy-Paste Is OK**: Sometimes duplication beats the wrong abstraction.

## The Check

Ask yourself:
- How many levels deep is my action chain?
- Can I test this action in isolation?
- Would debugging require traversing 5+ files?
- Am I abstracting because I should, or because I can?

## The Protocol

1. **Start Simple**: One action, one task.
2. **Extract When Repeated**: Second time you need logic, consider extraction.
3. **Compose Don't Inherit**: Actions call actions, not extend them.
4. **Flatten When Deep**: If chain > 4 levels, rethink the design.

## Real-World Examples

See [examples.md](examples.md).