How to use the Corporate Event Analysis agent skill

What the skill makes the agent do

The skill's source file is written in Chinese and describes event-driven analysis of company announcements. Its core idea is that announcements contain new information, the market takes time to absorb it, and returns around the event can differ from normal. The file lists five areas: merger and restructuring arbitrage, the signals in major shareholder and executive share changes, the reading of equity incentive plans, the effect of private placements and rights issues, and the avoidance or speculation around ST and delisting risk.

The file provides reference tables, formulas and a report template. For example, it gives a formula for a merger arbitrage spread, a table of how it rates insider share-change signals, checklists for judging an incentive plan, and a scoring idea for how a private placement's use of funds might be viewed. It does not fetch live data. The agent works from the announcement text and figures you give it, applying the file's framework.

Install the corporate-events skill

The skill is distributed as a ZIP file from the directory page. Download the ZIP and unzip it. The skill folder contains a file named SKILL.md, which is the only file listed for this skill. Place the unzipped folder in one of two locations: a personal skills folder at ~/.claude/skills/, which makes it available across your projects, or a project folder at .claude/skills/ inside a specific project, which limits it to that project.

If you use the Claude apps instead of a local folder, you can upload the skill under Settings → Capabilities → Skills.

• Personal install: put the unzipped folder in ~/.claude/skills/
• Project install: put the unzipped folder in .claude/skills/ within your project
• Claude apps: upload the skill under Settings → Capabilities → Skills

Install the Python dependencies

The file's dependencies section lists two Python packages, pandas and numpy. Install them with the command below. The file's own code does not require any data API key.

The file also mentions the tushare data API as an optional source for insider-trade, incentive and placement data. The notes say it provides those interfaces, but the file does not require a key in its own code. Treat tushare as something you would set up yourself if you want that data; the skill does not depend on it to give guidance.

Trigger the skill with a natural request

You do not need a special command. The skill's description tells the agent when it applies: company events such as takeovers, major shareholder share changes, equity incentives, private placements and ST or delisting warnings on A-shares. Ask about one of those events and name the company or stock code, and the agent can load the framework.

Give the agent the announcement text or key facts from it, because the skill does not look anything up. The more specific your request, the more useful the structured output will be. Ask for the signal strength, the spread, or a report in the file's format.

• Example request: Analyse this controlling shareholder share-increase plan for a listed A-share company, rate the signal strength, and give an event report. Then paste the announcement.
• Example request: Calculate the merger arbitrage spread for this tender offer at 25 yuan with the stock at 23.5 yuan, assuming about three months to completion.
• Example request: Check this company's ST risk using these figures and tell me which warning category applies under the file's rules.
• Example request: Review this equity incentive plan and list its exercise price discount, vesting conditions and share source.

Read the output the skill produces

For event-driven reports, the file provides a template with four parts: an event summary (security, event type, date and scale), a signal assessment (strength, historical reference, certainty), a strategy suggestion, and risk notes. The strategy part can include entry timing, position size, holding period, stop-loss and take-profit levels.

The file's example figures are illustrative. The file itself says some of its historical return figures are unsourced, so check any number the agent gives you against an original source before relying on it.

• Event summary: security, event type, date and size
• Signal assessment: strength, historical reference and certainty
• Strategy suggestion: entry, position size, holding period, stop-loss and take-profit
• Risk notes: market-wide falls, slow execution of the event and sector valuation

Safety notes before you use it

The skill runs no scripts and makes no network requests. It is text guidance and reference tables. The main risks come from how the content is used. The file presents its strategies and position sizes as trading advice, and it states that the historical return figures are unsourced. It also warns that announcement data from third-party platforms lags, that the official exchange sites publish first, and that trading just before an announcement can be insider trading. It says event-driven strategies should keep single-event positions under 10%.

• No scripts are run and no network access is used
• Returns and position sizes in the file are presented as trading advice and some returns are unsourced
• Third-party data lags, so an arbitrage window may already be closed
• Trading around announcements carries insider-trading risk and possible regulatory scrutiny
• The rules and thresholds are specific to A-shares

Tips for better results

Paste the full announcement text rather than a summary, so the agent can check the terms such as the price, share count, vesting conditions and timing. Ask the agent to state which parts of the framework it applied, so you can check its reasoning. Keep the source announcement open next to the output, because the skill cannot confirm a fact that is not in your input.

Note that the file is written in Chinese. You can ask in English, but check that terms such as the warning categories and the stock exchange rules are translated correctly.

• Paste the full announcement text
• Ask which framework rules the agent applied
• Check each figure against the original announcement
• Expect Chinese terms from the source file