How to use the Financial Regulation Knowledge Base agent skill

What does the skill make the agent do?

An agent skill is a folder with a SKILL.md file that tells the agent when and how to use a body of knowledge. This one holds market rules written as a guide. The file says that wrong rule assumptions can distort backtests, for example a China A-share backtest that ignores daily price limits, or cause real-world breaches such as the US PDT restriction.

When you ask the agent to build or test a strategy, it can apply these constraints so the backtest does not assume trades that a real market would block. The skill is reference knowledge only. It contains no market data and does not fetch anything live.

• A-share: daily price limits by board, T+1 settlement, margin and short-borrow rules, call auction timing, block trades
• Hong Kong: T+0 trading, no daily price limit, short-selling rules, T+2 settlement
• US: PDT rule, circuit breakers, LULD pauses, RegSHO short-sale rules
• Crypto: 24/7 trading, regulatory status by region, leverage and liquidation notes
• Cross-border tax basics and a cost-checklist for multi-market strategies

How do you install the regulatory-knowledge skill?

Download the ZIP file from the skill's page on this directory. Unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder if you want the skill available only in that project. The skill's folder should contain SKILL.md.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there. Installation does not require running any script. The only optional dependency is Python with pandas and numpy, which you need only if the agent writes code that uses them.

How do you trigger the skill with a natural request?

You do not need a special command. Describe a task that involves trading rules, and the agent can use the skill when the request matches its description: financial regulation for A-shares, Hong Kong, US and crypto markets. Naming the market and the rule you care about makes the match more likely.

Put the constraint in the request itself. A request that says the backtest must respect A-share price limits and T+1 settlement gives the agent a clearer target than a generic request to backtest a strategy.

• Build a backtest for a moving-average strategy on A-shares, and respect price limits and T+1 settlement.
• Check whether this long-short pair trade between A-shares and Hong Kong stocks has hidden costs such as stamp duty and short-borrow fees.
• Add PDT limits to my US day-trading backtest and flag any trades that would breach them.
• Which crypto regulatory issues should I account for when backtesting a 24/7 strategy?

What examples show the skill in use?

The file includes an example of a compliance report for a pair trade between a mainland A-share and a Hong Kong stock. The report lists rule constraints, a cost estimate for each market, and suggestions. It shows the kind of output to expect: a structured checklist rather than a single number.

The cost checklist in the file lists items to check before you trust a cross-market backtest: overlapping trading hours and time zones, differing holidays, currency risk, whether short signals can be executed, minimum lot sizes, and implicit costs such as stamp duty and fees.

• Price-limit handling: a buy signal at a limit-up price may not fill, so the fill is pushed to the next day.
• Settlement timing: a same-day sell after a buy is flagged as overstating returns under T+1.
• Short-borrow cost: the file gives an annualised A-share short-borrow fee of 8 to 10 percent, which a short strategy must subtract.
• Minimum lot size: A-shares trade in 100-share lots, which can make exact position sizing hard with small capital.

What does the skill need, and what should you know about safety?

The skill's listed requirements are Python with pandas and numpy, needed only when the agent writes code that uses them. The skill does not run scripts and does not use the network, according to its description. It consists of the single file SKILL.md.

Because the skill makes no network calls, it cannot check current fees, margin ratios or policy status. The file itself says its rules are summarised from memory, change often, and must be checked against the date and the broker. Treat its figures as starting assumptions.

• Runs scripts: no
• Uses network: no
• Files: SKILL.md
• Licence: MIT

What tips help you get reliable results?

Ask the agent to state which rule version or date it applied. The file warns that rules change, so backtests should use the rules that applied at each historical point. Also ask for conservative values where brokers differ, because the file notes that commission, short-borrow fees and latency vary widely between brokers.

Keep the skill's output as a checklist for your own review. Verify any figure that affects real money or a tax filing with a professional and the broker.

• Name the market and the date range of the backtest in the request.
• Ask for the cost and tax effects to be shown separately, so you can remove them to test sensitivity.
• Use the file's rule constraint matrix to compare markets side by side.
• Check the figures the agent quotes against your broker's current terms.