The Trading Development Kit

Write your system once. Stop donating to the market.

A Claude Code skill that scaffolds a complete, five-layer rule-based trading operation. Every layer has one job. Together they remove discretion from your process and replace it with a system you can trust.

---

The Five Layers

Layer	Component	Job
L1	CLAUDE.md / Risk.md	Memory — Trading rules, risk parameters, session behavior. Always loaded.
L2	Playbooks / Skills	Knowledge — Setup library: breakout, pullback, mean-reversion. On-demand.
L3	Hooks	Guardrail — Shell scripts that fire automatically around session events.
L4	Subagents	Delegation — Specialist agents run in their own context and return one clean answer.
L5	Plugins	Distribution — Bundle your entire system into a deployable package.

---

Install

Option 1 — Install the .skill file directly

Download the-trading-dev-kit.skill from the Releases page and install via Claude Code:

claude skill install the-trading-dev-kit.skill

Option 2 — Clone and use locally

git clone https://github.com/<YOUR_USERNAME>/the-trading-dev-kit
cd the-trading-dev-kit

Then add the skill to your Claude Code project:

cp -r the-trading-dev-kit ~/.claude/skills/

---

Quick Start

Once installed, trigger the skill with any of these prompts in Claude Code:

Full system setup:

Set up my full trading system. I trade momentum breakouts on the 5-minute chart.
My max daily loss is $500 and I risk 1% per trade.

Build Layer 1 (rules + risk):

Build my CLAUDE.md for trading. I trade SPY options on the 5-minute timeframe.
My edge is VWAP reclaims. Max daily loss $300. Risk 1% per trade.

Write a playbook (Layer 2):

Write a breakout.md playbook for ES futures. Include volume confirmation rules,
entry trigger, stop placement, scaling logic, and invalidation conditions.

Set up hooks (Layer 3):

Write a PreMarket.sh hook that loads the overnight range, sets a session bias,
and outputs a briefing in under 60 seconds.

Deploy a subagent (Layer 4):

Spawn risk-manager subagent. Account: $25,000. Trade: AAPL long at $195,
stop at $193. Return recommended size and downside scenario.

Package for distribution (Layer 5):

Build a plugin.json manifest for my trading system and package it for team distribution.

---

Repository Structure

the-trading-dev-kit/
├── SKILL.md                        ← Main skill file (orchestration)
└── references/
    ├── layer1-memory.md            ← CLAUDE.md + Risk.md templates
    ├── layer2-knowledge.md         ← Playbook templates (breakout, pullback, mean-reversion)
    ├── layer3-hooks.md             ← Hook scripts (PreMarket, PostTrade, EndOfDay)
    ├── layer4-subagents.md         ← Subagent definitions (market-researcher, risk-manager, journal-analyzer)
    └── layer5-plugins.md           ← Plugin manifest + packaging instructions

---

What Each Reference File Contains

layer1-memory.md

• Full CLAUDE.md template with trading.rules, risk.rules, execution.workflow, session.behavior
• Risk.md template with position sizing formula, drawdown limits, stop-loss rules
• Risk.local.md template for private account overrides (gitignored, never committed)
• Starter prompts: build, enforce, debrief

layer2-knowledge.md

• breakout.md template: valid structure, volume confirmation, entry, stop, scaling, invalidation
• pullback.md template: trend definition, acceptable depth, structure confirmation, exit plan
• mean-reversion.md template: extension conditions, fade entry trigger, hard stop, profit target logic
• Starter prompts: write, load, compare, build from history

layer3-hooks.md

• PreMarket.sh: overnight range, market regime, key levels, session bias, scheduled events, risk frame
• PostTrade.sh: instrument, direction, setup type, entry, stop, target, size, fill, risk dollars
• EndOfDay.sh: total trades, P&L, rule adherence score, session summary
• Claude Code settings.json hook configuration example

layer4-subagents.md

• market-researcher.md: catalysts, market structure, scheduled events, session bias
• risk-manager.md: position sizing, risk budget, downside scenarios, trade recommendation
• journal-analyzer.md: statistics, recurring mistakes, setup performance, rule violations, key change
• Starter prompts for each subagent

layer5-plugins.md

• plugin.json manifest template with all four component sections
• Full recommended repository structure
• Pre-ship audit checklist (12 items)
• Packaging and publishing workflow (bundle → commit → release → install)
• Starter prompts: build manifest, package, audit, version bump

---

The Core Principle

Rules over feelings. If it is not written here, it does not exist in your system. A good setup executed inconsistently is worse than no setup at all — it gives you false data about what works. This kit removes inconsistency from the equation.

---

License

MIT — use freely, adapt to your system, and share with your team.