Skip to main content
skill CX Codex

all-plan — Claude Code skill: what it does, how to install

all-plan runs multi-perspective planning for ambiguous tasks using designer, creative brainstorming, and scored review before execution.

no license ★ 475 updated 2026-10-13

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add cnfjlhj/ai-collab-playbook --skill all-plan -a claude-code
Open on GitHub

What all-plan does

This skill structures messy assignments by simulating distinct viewpoints before any code gets written. It begins by clarifying requirements across five dimensions. A dedicated inspiration stage gathers creative and aesthetic concepts, followed by the designer role compiling a complete blueprint independently.

A separate reviewer persona grades the resulting plan against an integrated rubric. The blueprint must reach a score of at least 7.0 to pass. If it falls short, an automated correction cycle triggers for at most 3 rounds. Without reaching that 7.0 threshold within 3 attempts, the plan fails inspection.

What it helps with

  • draft implementation steps for poorly defined tasks
  • collect creative concepts before building technical specs
  • evaluate architecture proposals against a scoring rubric
  • refine draft blueprints automatically across revision rounds

Example prompts for the agent

  • Run all-plan to design the notification pipeline with vague requirements
  • Draft an implementation blueprint for the user dashboard overhaul using all-plan
  • Plan the checkout refactor with multi-role review via all-plan

How to install all-plan

1Skills CLI (Vercel)

npx skills add cnfjlhj/ai-collab-playbook --skill all-plan -a claude-code

2Manually: put the files in place

~/.claude/skills/all-plan/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/all-plan/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about all-plan

How do I install all-plan in Claude Code?

Run npx skills add cnfjlhj/ai-collab-playbook --skill all-plan -a claude-code in your terminal. Alternatively, install it by placing the definition at ~/.claude/skills/all-plan/SKILL.md.

What does all-plan do?

It orchestrates requirement clarification across five angles, creative brainstorming, architecture design, and a scored review loop before writing code.

What score is required to pass the review?

The reviewer must award a score of at least 7.0. If the draft fails, the auto-correction loop runs up to 3 times to fix issues.

Similar skills

All skills for Agent setup: memory, context, orchestration →

From the same set18

The whole set →
提示词优化器 提示词优化器 is a Claude Code prompt that turns a vague request into a usable instruction through questions and revisions. It clarifies requirements before asking the agent to act. GP ChatGPT ★ 475repo stars skill-governance-loop skill-governance-loop — a skill for managing other skills: checks quality, compares versions, and decides whether to keep, enable, hide, disable, or remove a given skill from your local collection. CX Codex +1 ★ 475repo stars prompt-polisher prompt-polisher converts rough notes and stream-of-consciousness transcriptions into clean, reusable prompts. CX Codex +1 ★ 475repo stars find-skills find-skills is a skill that searches the registry and, if needed, GitHub for reusable skills. It helps check for an existing workflow before building one. CX Codex +1 ★ 475repo stars skills-governance skills-governance scans local skill folders, groups duplicates by name, and sorts skills into keep, disable, or archive. CX Codex +1 ★ 475repo stars xhs-note-creator xhs-note-creator generates complete Xiaohongshu note packages with copy, card Markdown, rendered images, and publishing automation. CX Codex +1 ★ 475repo stars proactive-explorer The proactive-explorer skill tells Claude Code to look for missing context in the project before asking the user. Use it when code, configuration, or logs may resolve an ambiguity. CX Codex +1 ★ 475repo stars paperreview Uploads a paper PDF to paperreview.ai over HTTP and fetches the finished review by token. No browser in the loop. CX Codex +1 ★ 475repo stars

Project sponsors

Companies that keep this analytics open