#312312th of 324 in Skills, plugins and prompts

paul

Plan-Apply-Unify workflow commands for structured development in Claude Code

Stars
1.3k
License
MIT
Last commit
Aug 2026
Last release
Mar 2026
Language
JavaScript
Stars, last 30 weeks+634
3 Apr 2026: 626 stars10 Apr 2026: 682 stars17 Apr 2026: 737 stars24 Apr 2026: 774 stars1 May 2026: 828 stars8 May 2026: 865 stars15 May 2026: 898 stars22 May 2026: 907 stars29 May 2026: 925 stars5 Jun 2026: 942 stars12 Jun 2026: 951 stars19 Jun 2026: 976 stars26 Jun 2026: 1,016 stars3 Jul 2026: 1,050 stars10 Jul 2026: 1,079 stars17 Jul 2026: 1,100 stars24 Jul 2026: 1,122 stars31 Jul 2026: 1,132 stars7 Aug 2026: 1,141 stars14 Aug 2026: 1,150 stars21 Aug 2026: 1,175 stars28 Aug 2026: 1,193 stars4 Sept 2026: 1,204 stars11 Sept 2026: 1,215 stars18 Sept 2026: 1,225 stars25 Sept 2026: 1,236 stars2 Oct 2026: 1,240 stars9 Oct 2026: 1,246 stars16 Oct 2026: 1,250 stars23 Oct 2026: 1,254 stars30 Oct 2026: 1,260 stars

626 on 3 Apr 20261,260 on 30 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

PAUL installs 26 slash commands into Claude Code that enforce a plan, apply, unify loop. Plans carry Given/When/Then acceptance criteria, each task is checked against the spec after execution, and project state lives in a .paul/ directory (STATE.md, ROADMAP.md, per-phase PLAN and SUMMARY files). Implementation stays in-session; subagents are used only for research and discovery.

Who it is for: Claude Code users who want a structured plan-and-close workflow

Strengths

  • State persists in plain markdown files under .paul/, so sessions resume with /paul:resume
  • Each task gets an Execute/Qualify check against its acceptance criteria, with four status codes
  • Installs with one npx command, globally or per project, on Mac, Windows and Linux
  • Scope-adaptive plans (quick-fix, standard, complex) with a coherence check before approval

Weaknesses

  • Works only with Claude Code; no other coding agent is mentioned
  • Graph, cost attribution and rule injection need BASE v2, linked from a community site
  • Mandatory UNIFY step and per-task qualification add process overhead to small changes
  • Claims such as subagent output being ~70% quality are stated without benchmarks

What it needs

  • no GPU
  • Needs Claude Code, Node.js (npx)
  • Models: Claude

Quick answers

Is paul open source?

Yes. Its code is on GitHub under the MIT license.

Also in Skills, plugins and prompts

See all 324
Also in Skills, plugins and prompts
RankProjectScore
1impeccableDesign skill with 24 slash commands for AI coding agents' frontend output Live demo ↗ (opens in a new tab)98.9k starsApache-2.096 out of 100
2skillsAgent skills for planning, TDD, debugging and code review308.9k starsMIT95 out of 100
3eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses294.5k starsMIT95 out of 100
#4gstackMarkdown slash-command skills that give Claude Code planning, review, QA and ship roles138.5k starsMIT95 out of 100
#5cavemanCuts token use in coding agents with terse replies and compressed inputs Live demo ↗ (opens in a new tab)116.6k starsApache-2.095 out of 100