Guide · Agent skill
How to use the Harmonic Patterns agent skill (harmonic)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The harmonic skill teaches an AI agent to look for five-point XABCD price shapes such as Gartley, Bat, Butterfly and Crab, using Fibonacci ratios to define them. It is written for traders and analysts who already know harmonic pattern theory, developers building screening or backtesting prototypes, and students studying how these rules are coded. After this guide you will know what the skill does, how to install it, how to phrase a request so the agent uses it, and what to check before relying on its output.
Harmonic Patterns Signal EngineHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Download the ZIP. Get the harmonic skill ZIP from the directory page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use or into .claude/skills/ inside a project.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill.
- 4Install the requirements. Make sure Python with numpy and pandas is available. Optionally run pip install "vibe-trading-ai[harmonic]" to add pyharmonics.
- 5Supply your price data. Provide historical OHLC price data yourself, since the skill does not fetch it.
- 6Ask in plain language. Request a scan or check using terms like XABCD, Gartley, Bat, Butterfly, Crab or PRZ, and set is_stock if the data is a stock.
- 7Review the signals. Check the 1, -1 or 0 output against the reference notes and your own validation before acting on any of it.
What the harmonic skill makes the agent do
An agent skill is a folder of instructions and optional scripts that an AI agent loads when a task matches the skill's description. This one covers harmonic patterns. Each pattern is a five-point structure labelled X, A, B, C and D, where the distance between points follows fixed Fibonacci ratios.
The skill describes four patterns. Gartley retraces to 0.618 of XA at point B and 0.786 of XA at point D. Bat uses 0.382 to 0.5 of XA at B and 0.886 of XA at D. Butterfly uses 0.786 of XA at B and 1.27 of XA at D, which is the only one where D extends beyond X. Crab uses 0.382 to 0.618 of XA at B and 1.618 of XA at D, the most extended of the four.
Understand the PRZ and the signal convention
The PRZ (Potential Reversal Zone) is the area around point D where several Fibonacci projections converge. The skill treats this area as the place where a reversal is more likely. A bullish pattern has D at the bottom and maps to a buy signal. A bearish pattern has D at the top and maps to a sell signal.
The signal is encoded as 1 for long, -1 for short and 0 to stand aside. The skill also notes that a swing point is only known a few bars after it forms, so the bundled detector stamps the signal on the confirmation bar rather than on the swing bar itself.
- 1 = long
- -1 = short
- 0 = stand aside
Install the skill in Claude
Download the ZIP for the harmonic skill from the directory page. Unzip it into a skills folder. For a personal setup, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.
If you use the Claude apps instead of Claude Code, you can upload the skill under Settings → Capabilities → Skills.
The skill's own dependency note gives one optional install command for the pyharmonics library. Install it only if you want that path.
- Personal: ~/.claude/skills/
- Per project: .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills
pip install "vibe-trading-ai[harmonic]"Install the requirements
The skill needs Python with numpy and pandas. Historical price data must be supplied by you; the skill does not fetch it. The pyharmonics library is optional. If the extra is not installed, the example signal engine falls back to the bundled detector, so the skill still runs without it.
- Python with numpy and pandas (required)
- Your own historical price data (required)
- pyharmonics via the optional extra (optional)
Trigger the skill with a natural request
You do not need a special command. Describe the task in plain language that mentions harmonic patterns, XABCD, Gartley, Bat, Butterfly, Crab or PRZ, and attach or point to your price data. The agent matches your request against the skill's description and loads the instructions.
Be specific about the instrument and the output you want. The skill has one documented parameter, is_stock, which defaults to False and changes the analysis parameters when the instrument is a stock. Say so if you are working with stock data.
- Scan this OHLC file for Gartley and Bat patterns and list each XABCD point with its date.
- Check the last 500 bars for bullish Crab patterns and report where D sits relative to the PRZ.
- Run the harmonic detector on this stock series with is_stock set to true and output 1, -1 or 0 per bar.
Read the output and the reference notes
The bundled references explain the rules behind the signals. The PRZ reference describes possible entry, stop and target logic: wait for price to react in the zone with a reversal candle rather than entering before D is confirmed; place the stop beyond D; and use the 0.382 and 0.618 retracements of the AD leg and point A as targets. The pattern reference lists the ratio rules for each shape.
These notes also say that completing a pattern does not guarantee a reversal, that volume and candle confirmation are needed, and that higher timeframe patterns are considered more reliable. Treat them as the author's rules to test, not as proven results.
- references/PRZ交易.md covers PRZ entry, stop and target rules
- references/XABCD形态.md covers the ratio rules for each pattern
Stay within the safety limits
The skill runs scripts and needs no network access, according to its listing. It works on data you provide. Review the bundled example_signal_engine.py before running it on any data that matters, because its signal logic and parameters have not been validated in the files provided. Do not connect its output to a live trading account without independent review.
- Runs scripts: yes (review example_signal_engine.py first)
- Uses network: no
- Signal logic is not validated in the files provided
What people use it for
- Screening a historical price file for Gartley, Bat, Butterfly and Crab shapes and listing each XABCD point.
- Building a backtesting prototype that converts pattern completions into 1, -1 or 0 signals per bar.
- Studying how Fibonacci ratio rules for each harmonic pattern are written in code.
- Checking whether recent price action has reached a PRZ and what the reference notes suggest about entry and stops.
Frequently asked questions
Is the harmonic pattern skill free?
The source lists the skill under an MIT licence, which governs use of the code. The source does not state any price, so check the directory page for current terms. The optional pyharmonics library is installed separately with pip.
What do I need to run the harmonic skill?
You need Python with numpy and pandas, plus your own historical price data, since the skill does not supply data. The pyharmonics library is optional. Without it, the example engine uses its bundled detector.
How do I install a Claude agent skill from a ZIP?
Download the ZIP and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ inside a project. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.
Does the harmonic skill predict market moves accurately?
No evidence of predictive accuracy is given in the source. The skill produces mechanical pattern matches and reversal signals, and markets can move against them. Test any output on your own data before relying on it.
How does the harmonic skill compare to other pattern tools?
The source does not compare this skill with other tools, so no comparison can be claimed here. Its distinguishing features are the four XABCD patterns it defines, the PRZ concept, and the 1, -1, 0 signal convention, all documented in its SKILL.md and reference files.
Why does my harmonic signal appear a few bars after the swing point?
The skill states that a swing point is only known swing_window bars after it forms. The bundled detector therefore stamps the signal on that confirmation bar, not on the swing bar. The optional pyharmonics path is unchanged.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits traders who already use Fibonacci harmonic theory, developers prototyping screens or backtests, and students learning how these rules are coded. It is not suited to anyone expecting reliable profits or personal financial advice, and its signal logic should be validated before any real-money use.
Download the skill