Guide · Agent skill
How to use the Shadow Account trade analyzer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Shadow Account is an agent skill from the Vibe-Trading project that reads your uploaded trade journal, a brokerage statement, and tries to describe how your profitable trades usually look. It then backtests that pattern, compares the result with what you actually earned, and explains the gap. This guide covers what the skill does, how to install it, how to ask for it in plain language, and what to watch for. Everything here comes from the skill's SKILL.md file.
Shadow Account Trade AnalyzerHKUDS/Vibe-Trading · MIT Download the skill Quick start
- 1Get the ZIP. Download the skill ZIP from this page. It contains SKILL.md.
- 2Install the skill. Unzip into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 3Prepare your journal. Upload your brokerage statement as the trade journal and make sure analyze_trade_journal has run.
- 4Ask for extraction. Request something like extracting your strategy and review the 3-5 rules the agent proposes.
- 5Confirm or adjust the rules. Say whether the rules sound like you. If not, the agent raises min_support and extracts again.
- 6Run the backtest. The agent backtests the rules across China A-shares, Hong Kong, US stocks and crypto and computes the P&L difference.
- 7Read the report. Review the attribution breakdown and the 8-section PDF, or the HTML if PDF generation failed.
What is an agent skill, and what does Shadow Account do?
An agent skill is a folder containing a SKILL.md file that tells an AI agent when to use a specific workflow and which tools to call. Shadow Account is one such skill, named shadow-account, in the analysis category.
The skill runs a four-step workflow. It extracts 3-5 plain-language rules from your profitable round-trips, backtests those rules across China A-shares, Hong Kong, US stocks and crypto, breaks down the difference between the shadow result and your real result, and renders a report. An optional step lists symbols that currently fall in the rule entry pattern.
- Extracts rules such as human-readable descriptions with a support count, coverage rate and holding-day range
- Backtests per market and in a combined pool, reporting Sharpe, annualised return and maximum drawdown
- Attributes the P&L gap to noise trades, early exits, late exits, overtrading and missed signals
- Produces HTML and PDF output, with the PDF following an 8-section layout
What do you need before you start?
The skill needs a Vibe-Trading agent environment that provides five tools: analyze_trade_journal, extract_shadow_strategy, run_shadow_backtest, render_shadow_report and scan_shadow_signals. You also need a trade journal file uploaded by the user. The skill expects analyze_trade_journal to have been run already, and if it has not, the agent should run that tool first.
Two extras are optional. weasyprint enables PDF output; if it fails, the skill falls back to HTML only. An llm_translator callable can turn the rule conditions into more natural Chinese sentences; without it, a built-in template is used.
- Required: Vibe-Trading agent environment with the five named tools
- Required: a user-uploaded trade journal (brokerage statement)
- Optional: weasyprint for PDF output
- Optional: an llm_translator callable for natural-language rule text
- The skill is marked as running scripts and not using the network
How do you install the skill?
Download the ZIP file from this page. It contains the SKILL.md file. Then choose one of the two installation routes below.
For Claude Code, unzip the file into ~/.claude/skills/ to make it available to you across projects, or into a project's .claude/skills/ folder to limit it to that project. For the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: upload the ZIP under Settings → Capabilities → Skills
How do you trigger the skill with a natural request?
You do not need a special command. The skill loads when you make requests like the ones it lists as triggers. These include asking to extract your strategy, train your shadow, backtest your approach, see how much more you could have earned, or describe your profit pattern.
Make sure your journal has been uploaded first. Then ask in plain language and let the agent handle the steps in order.
- "Extract my strategy from my trade journal"
- "Train my shadow on this statement"
- "Backtest my approach and tell me how much more I could have made"
- "What is my profit pattern?"
What does a typical run look like?
The agent first asks you to confirm the extracted rules, for example whether they look like your own way of trading. If you say they do not, the skill raises min_support and re-runs extraction. After confirmation, it backtests across the four markets by default and renders the report.
The most useful output is the attribution section. It shows the shadow P&L, your real P&L and the difference, with the noise trades (real trades that match none of your profitable rules) called out. Every figure is signed, so a positive value means the shadow did better.
- Rule cards: rule_id, human_text (30 characters or fewer), support_count, coverage_rate, holding_days_range
- Backtest matrix: per_market metrics, combined pool metrics and an equity curve
- Counterfactual top 5: the trades most worth doing or avoiding, with dates and reasons, ranked by impact
What safety limits should you know about?
The skill states that it never connects to any order channel, so it cannot place trades. Its scan output is for research only and is not a buy recommendation. Backtests on past trades do not predict future returns.
The skill also refuses to invent rules. If the journal has fewer than 5 profitable round-trips, it raises an error. It extracts rules only from your own trades, not from community or public strategies. The metadata marks it as running scripts, so review the files before installing it in a sensitive environment.
- No order placement: the tools never connect to a trading channel
- Scan output is research only, not advice
- Needs at least 5 profitable round-trips or it stops with an error
- Uses only the user's own journal, not outside strategies
What tips help you get better results?
Upload a complete journal so the rules have enough trades to support them. If the first rules do not feel like you, ask the agent to raise min_support and run extraction again rather than accepting weak patterns.
Read the attribution section with the noise trades in mind. It is the part of the report that shows where emotion, rather than your stated rules, shaped results.
- Use a full journal rather than a short excerpt
- Reject rules that do not sound like you and ask for a re-run with a higher min_support
- Treat the scan list as a prompt for further research, not a trade signal
What people use it for
- Checking whether your profits come from a repeatable pattern or from a few lucky trades
- Finding how much P&L was lost to trades that broke none of your profitable rules
- Spotting exits that came too early or too late relative to your typical holding period
- Giving an analyst a structured backtest and attribution breakdown of a trader's journal
- Building a trading-review workflow on top of the Vibe-Trading toolset
Frequently asked questions
How do I install the Shadow Account skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills instead.
What do I need to run the shadow-account skill?
You need a Vibe-Trading agent environment that provides the analyze_trade_journal, extract_shadow_strategy, run_shadow_backtest, render_shadow_report and scan_shadow_signals tools, plus a user-uploaded trade journal file. weasyprint is optional for PDF output.
Is the Shadow Account skill free and what is its licence?
The item is listed under the MIT licence. This page does not state any price, so check the source repository for any terms beyond the licence.
Can Shadow Account place trades for me?
No. The skill states that it never connects to any order channel and its scan output is for research only, not a buy signal.
Why does the skill say I need at least 5 profitable round-trips?
The skill is designed to raise an error when the sample is too small, rather than invent rules. With fewer than 5 profitable round-trips in your journal, it will stop instead of producing a strategy.
What if the PDF report does not generate?
The render step falls back to HTML-only output if weasyprint fails. You will still get the report in HTML form with the same key figures.
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
Shadow Account suits retail traders who want to test whether their own repeatable habits explain their profits, and analysts who need a structured attribution of a journal. It is a poor fit if you want trade advice, buy signals or automated order placement, since the skill is explicitly research-only and never trades. If your journal has fewer than five profitable round-trips, choose a different tool or add more history first.
Download the skill