# Pdsa Improvement Loop

> Run a PDSA continuous improvement loop: plan, do, study, act. Use when Codex is asked to improve a process with small experiments, learning cycles, operational changes, service improvements, or quality improvement.

- Skill: `walterkken/pdsa-improvement-loop` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add walterkken/pdsa-improvement-loop`
- Raw SKILL.md: https://api.skillmd.com/api/skills/walterkken/pdsa-improvement-loop/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: walterkken (https://skillmd.com/u/walterkken)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/walterkken/pdsa-improvement-loop

---


# PDSA Improvement Loop

## Workflow

1. Plan: define the problem, prediction, test, metric, scope, and timebox.
2. Do: run the smallest useful experiment.
3. Study: compare results with prediction and inspect unexpected effects.
4. Act: adopt, adapt, abandon, or run another cycle.
5. Record learning and next cycle.

## 马斯克 Supervision

Delete unnecessary process steps before running improvement experiments. Prefer small tests that reveal truth fast.

## Output

- problem statement
- prediction
- test plan
- metric
- results
- learning
- next action

See `references/source-notes.md`.