Agent Skills

PyBroker ships Agent Skills that give coding agents workflows for writing strategies, indicators, models, and backtests.

Installation

Public skills live under the skills/ directory in the PyBroker Git repository. Each skill is contained in its own folder and defined by a standard SKILL.md file.

The recommended way to install them is the Skills CLI, which supports Claude Code, OpenAI Codex, Cursor, and many other coding agents:

# Inside your own working project directory:
npx skills add edtechre/pybroker

The command asks which skills to install and which agents to install them for, then copies each one into that agent’s skills folder. Add --all to install every skill for every detected agent without prompting.

Manual Install

You can also clone the PyBroker repository and symlink the skills yourself:

git clone https://github.com/edtechre/pybroker.git

# Inside your own working project directory:
mkdir -p .claude/skills

# Symlink all PyBroker skills into your local Claude config
for skill in /path/to/pybroker/skills/*; do \
    ln -s "$skill" .claude/skills/; \
done

Replace .claude/skills with the folder that your own agent reads.

Claude Code

Claude Code discovers skills placed inside a .claude/skills/ folder in your project root. Once they are installed, every skill and its respective commands become available in the terminal.

OpenAI Codex

OpenAI Codex discovers skills in .agents/skills/.

Cursor

Cursor discovers skills in .cursor/skills/.

Claude Agent SDK

The Claude Agent SDK discovers skills from the .claude/skills/ folder installed above, so it loads them on demand and invokes them automatically without any prompt assembly:

pip install claude-agent-sdk

import asyncio
import os

from claude_agent_sdk import ClaudeAgentOptions, query

options = ClaudeAgentOptions(
    cwd=os.getcwd(),
    setting_sources=["user", "project"],
    skills="all",
    allowed_tools=["Read", "Write", "Bash"],
)


async def main():
    async for message in query(
        prompt="Create a mean-reversion PyBroker strategy on AAPL",
        options=options,
    ):
        print(message)


asyncio.run(main())

Pydantic-AI

Pydantic-AI builds agents in pure Python. It does not search for skill directories automatically, so you point its Skills capability at the .agents/skills library installed above:

pip install "pydantic-ai-harness[skills]"

from pydantic_ai import Agent
from pydantic_ai_harness.skills import Skills

agent = Agent(
    'anthropic:claude-opus-5',
    capabilities=[Skills('.agents/skills')],
)

result = agent.run_sync("Create a mean-reversion strategy on AAPL")
print(result.output)

Available Skills

pybroker-strategy-creator

Create practical PyBroker strategy code from user intent while preserving backtest hygiene, including no lookahead leakage, explicit sizing, clear risk controls, and locally valid PyBroker API usage.

See full SKILL.md on GitHub

pybroker-indicator-creator

Write fast, correct PyBroker indicators by registering vectorized NumPy/Numba functions with pybroker.indicator, wiring them into strategy executions and models, and keeping every value free of lookahead bias. Covers the built-in indicator factories and vector helpers, custom Numba @njit kernels, wrapping third-party technical analysis libraries such as TA-Lib and pandas-ta, standalone computation with IndicatorSet, hyperparam-driven indicators, and multi-timeframe interval indicators.

See full SKILL.md on GitHub

pybroker-model-trainer

Wire machine learning models into PyBroker backtests by registering training and prediction functions with pybroker.model, feeding them indicator features, and evaluating them with walkforward analysis while keeping the train/test flow free of lookahead leakage. Covers per-symbol, pooled multi-symbol, per-bar time-series, lagged-feature, and pretrained models across common libraries such as scikit-learn, XGBoost, and arch.

See full SKILL.md on GitHub

pybroker-optimize

Tune PyBroker strategy hyperparameters with Strategy.optimize by declaring tunable values with pybroker.hyperparam, wiring them into indicators and execution functions, and scoring each candidate combination on a training window before the winning values are replayed on held-out test data. Covers grid, TPE, and random sampling through the integrated Optuna backend, custom Optuna samplers and studies, walkforward optimization across multiple windows, and reading results from OptimizeResult and the underlying optuna.Study.

See full SKILL.md on GitHub

pybroker-multi-interval

Build multi-timeframe PyBroker strategies that trade a base timeframe while confirming regime and trend on strictly coarser compressed intervals such as weekly and monthly bars. Covers the three interval formats, providing compressed bars with add_execution(intervals=...), computing indicators and training models per interval by binding them with .intervals(...), reading completed bars through ctx.interval(...), and standalone compression with compress_bars. Strategy code only ever sees completed compressed bars, which keeps higher-timeframe logic free of partial-bar lookahead.

See full SKILL.md on GitHub

pybroker-rotational-trading

Build rotational PyBroker strategies that hold the top-ranked symbols in a universe and rotate out names that fall from favor. Execution functions score symbols with ctx.long_score and ctx.short_score, cap positions with Strategy.set_max_long_positions and set_max_short_positions, and Strategy.enable_rotation(worst_rank_held=...) liquidates and refills slots each bar from the top-ranked candidates, optionally sized with a custom sizer over RotationContext. Also covers the simpler ranked-cap mode and dynamic universes via SymbolSelector.

See full SKILL.md on GitHub