Craft with Intention — Thoughtful skills for AI-driven development
Thinking Smart is a Claude Code plugin that scales your AI-driven workflows with structured planning, parallel execution, error recovery, and plan auditing. It transforms how Claude Code approaches complex tasks — from ad-hoc problem solving to intentional, reviewable engineering.
Website: nuoyimanaituling.github.io/thinking-smart
Install in Claude Code with two commands:
# 1. Add the marketplace
/plugin marketplace add nuoyimanaituling/thinking-smart
# 2. Install the plugin
/plugin install thinking-smart@thinking-smart| Skill | Command | Description |
|---|---|---|
| Brainstorming | /thinking-smart:brainstorming |
Explore user intent, requirements, and design before implementation. Emphasizes web search to verify ground truth and resolve conflicts by recency. |
| Write Plan | /thinking-smart:write-plan |
Create structured plan files with diagrams, before/after comparisons, and clear acceptance criteria — not walls of text. |
| Execute Plan | /thinking-smart:execute-plan |
Execute a plan file step by step, spawning parallel subagents for independent tasks to maximize throughput. |
| Audit Plan | /thinking-smart:audit-plan |
Verify planned tasks against actual implementation with evidence-based status: Done, Partial, or Missing. |
| Recover from Errors | /thinking-smart:recover-from-errors |
When tools fail, re-align with the plan file instead of guessing — preventing cascading errors and goal drift. |
| Using Skills | /thinking-smart:using-skills |
Skill discovery and invocation rules for Claude Code conversations. |
The plugin is built around five core principles that enhance Claude Code's capabilities through hooks and structured workflows:
- Ground Truth First — Web search during exploration to verify facts against authoritative sources, not just local code.
- Clear Plan — Plans use diagrams and structured templates instead of freeform text, making them easy to review at a glance.
- Parallelism — Independent tasks are automatically spawned as parallel subagents, maximizing concurrent execution.
- Plan Audit — After execution, every task is verified against the plan with evidence-based status reporting.
- Error Recovery — On tool failure, the agent consults the plan file and recovers on course, instead of guessing and drifting.
.claude-plugin/marketplace.json # Marketplace registry
claude/thinking-smart/
.claude-plugin/plugin.json # Plugin metadata (v1.0.x)
skills/
brainstorming/SKILL.md
write-plan/SKILL.md
execute-plan/SKILL.md
audit-plan/SKILL.md
recover-from-errors/SKILL.md
using-skills/SKILL.md
hooks/ # Claude Code event hooks
website.*.toml # Website content config
website/ # Astro static site
src/
public/images/philosophies/ # Generated illustration images
scripts/ # Content & image generation scripts
# Clone and configure Git hooks for automatic version bumping
git clone https://github.com/nuoyimanaituling/thinking-smart.git
cd thinking-smart
sh scripts/setup.sh-
Create the skill directory and definition:
mkdir -p claude/thinking-smart/skills/<skill-name> # Create claude/thinking-smart/skills/<skill-name>/SKILL.md
-
Test locally:
claude --plugin-dir ./claude/thinking-smart
Skills are namespaced as /<plugin-name>:<skill-name>.
The website/ directory contains an Astro static site showcasing the plugin.
cd website
npm install
npm run dev # Local dev server
npm run build # Build to dist/Pushes to master automatically deploy via GitHub Actions to GitHub Pages.
Made with intention by nuoyimanaituling.