🧠 Agent Skills Marketplace

Structured human+AI collaboration for knowledge work.

⚠️ Live experiment. My cognitive toolkit — fork it, adapt it to your brain.

💡 Why This Exists

AI coding tools are powerful but chaotic. Most people either micromanage every prompt or let AI run wild and pray. Neither scales to real work.

This toolkit treats AI collaboration as cognitive discipline — a set of thinking modes you activate depending on the situation, with the human always setting the pace.

🎯 Principles

🧭 Frame before act — classify the problem, then pick the right approach
🧠 Think before build — brainstorm, investigate, design before writing code
🚧 Human controls pace — gates between sections, you verify, AI continues
🪨 Boulder → Pebbles — scale your process to the problem (OpenSpec for big, just code for small)
💾 Sessions persist — save context, resume tomorrow without re-explaining
🪞 Learn from patterns — retrospectives extract what worked and what didn't

📖 Deep dive: PHILOSOPHY.md

🔄 Cognitive Modes

flowchart LR
    F["🧭 Frame"] --> T["🧠 Think"]
    T --> B["⚙️ Build"]
    B --> D["🔧 Debug"]
    D --> L["🪞 Learn"]
    L -.->|"next problem"| F

    classDef frame fill:#E8EAF6,stroke:#3F51B5,color:#000
    classDef think fill:#E1BEE7,stroke:#7B1FA2,color:#000
    classDef build fill:#C8E6C9,stroke:#388E3C,color:#000
    classDef debug fill:#FFE0B2,stroke:#F57C00,color:#000
    classDef learn fill:#BBDEFB,stroke:#1976D2,color:#000

    class F frame
    class T think
    class B build
    class D debug
    class L learn

Mode	What	Hero Skills
🧭 Frame	Classify the problem → route to the right skill chain	`/frame-problem` (Cynefin), `/pick-model`
🧠 Think	Divergent ideation, deep analysis, adversarial review	`/brainstorm` (SCAMPER), `/investigate` (Issue Trees, Pre-mortem), `/probe` (safe-to-fail), `/challenge` (Devil's Advocate)
⚙️ Build	Plan → develop → gate → test → sync	`/openspec-*` suite (7 skills, human-gated sections)
🔧 Debug	Search-first troubleshooting with learnings DB	`/troubleshoot` (Wolf Fence, 5 Whys, OODA), `/experiment` (Chaotic act-sense)
🪞 Learn	Extract patterns, persist sessions	`/retrospect-*`, `/save-context`, `/load-context`

Plus: tool creation (/edit-tool → routes to skill/command/agent editors), conversions (PDF, EPUB, Google Docs), and a GTD plugin for task capture.

🏪 Plugins

Plugin	Description	Status
dstoic	Core cognitive toolkit: 44 skills, 0 commands, 4 hooks	✅ v0.19.1
gtd	GTD workflow automation for Obsidian vaults	✅ v0.1.0
...	More planned	🔜

📦 Install

Install from the Claude Code marketplace:

/install-plugin https://github.com/digital-stoic-org/agent-skills

This installs all plugins (dstoic, gtd). To install a specific plugin only, add it to .claude/settings.json:

{"plugins": ["digital-stoic-org/agent-skills/dstoic"]}

🚀 Quick Start

/frame-problem how should I approach building a new auth system
# → Classifies as Complicated (governing constraints) + Boulder
# → Routes to: /investigate → /openspec-plan

/brainstorm product naming ideas for my CLI tool
# → Research → SCAMPER divergence → weighted scoring → recommendation

/troubleshoot "TypeError: Cannot read property 'map' of undefined"
# → WebSearch → qualify → diagnose → OODA → save learning

📖 Documentation Philosophy

GenAI makes it too easy to generate walls of text → cognitive overload for humans.

Doc Type	For	Example
📄 `README.md`	Humans (1 min scan)	This file
📚 `README-full.md`	Humans (deep dive)	Full catalog
🧭 `PHILOSOPHY.md`	Humans (mindset)	Principles
🤖 `SKILL.md`	LLMs (token-optimized)	Not for human reading

Respect your attention. Start here, dive deeper only when needed.

⚠️ 🪥 CLAUDE.md = toothbrush. See CLAUDE.md.example for inspiration, don't copy. (Why?)

The example uses rtk for token-optimized command output. Install it separately if you want to use the rtk instructions.

📚 Full skill catalog: README-full.md · 🧭 Philosophy: PHILOSOPHY.md

edit-command

SKILL.md