Skill Development Overview

Skills are packaged strategy modules for Orien trading agents. They define how an agent observes the market, plans candidate actions, checks risk, and executes approved order flow.

Skills are not the AI Chat tool extension layer. Use MCP Development when you want to add tools to chat.

When to Build a Skill

Build a Skill when you need:

repeatable trading strategy logic
factor combinations and signal generation
risk-aware execution behavior
backtestable and inspectable code
a reusable strategy package for pipeline or integrated agents

Skill Package Shape

A Skill package is typically structured like this:


my-skill/
├── SKILL.md
├── orien.runtime.yaml
└── scripts/
    └── strategy.py

SKILL.md describes the package. orien.runtime.yaml declares the runtime contract. scripts/ contains the code that the agent runtime calls.

Runtime Phases

Skills map tools into four runtime phases:

Phase	Purpose	Allowed Effects
observe	Gather market and account context	`read` only
plan	Produce a candidate action	`read` only
risk	Validate limits and reject unsafe actions	`read` only
execute	Submit approved order flow	`read` + `write`

The phase contract keeps analysis and execution separated. Observation and planning tools should not silently become execution paths.

Strategy Models

Python Skills

Python is the primary path for authoring strategy logic. It gives direct access to market data, analysis libraries, and exchange-facing tools.


async def get_market_data(symbol: str) -> dict:
    """Fetch market state for the current cycle."""
    async with httpx.AsyncClient() as client:
        response = await client.get(f"https://api.exchange.com/ticker/{symbol}")
        return response.json()

Rust Components

Rust can be used where deterministic behavior, lower-level control, or performance-sensitive logic is needed.

Prompt-Assisted Authoring

Prompting is useful for scaffolding and refinement. In the current architecture, its role is to help generate strategy structure, not to replace inspectable execution logic.

Product Fit

Well-designed Skills fit into the trading runtime:

Operator layer — Surfaces the strategy through consoles and reports
Strategy layer — Runs the Skill inside a structured agent cycle
Trading OS — Supplies data, execution, wallet, and risk primitives

MCP belongs to the AI Chat tool layer. Skills belong to the agent execution layer.

Next Steps

Skill Structure — Detailed breakdown of the package format
orien.runtime.yaml Reference — Runtime contract reference
Tool Development — Writing tool functions
Publishing — Submit to the Skills Market
MCP Development — Build tools for AI Chat