How to use the Earnings Revision & Guidance Analysis agent skill

What the earnings-revision skill makes the agent do

An agent skill is a folder containing a SKILL.md file with instructions the agent reads when a task matches the skill's description. This one holds a finance framework rather than software: formulas, thresholds and rules of thumb for judging an earnings report.

When it applies, the agent can judge whether analyst estimates are being revised up or down, how large a result was compared with expectations, and whether management's guidance language is positive or negative. It also explains post-earnings drift, the tendency of prices to keep moving in the direction of an earnings surprise for 60-90 days, and lists earnings-quality red flags such as cash flow lagging reported profit.

• Earnings surprise: standardized unexpected earnings (SUE) with interpretation bands
• Consensus revision breadth and estimate dispersion
• Guidance scoring against prior guidance and consensus, plus a table of guidance language patterns
• Pre-earnings and post-earnings checklists
• Earnings-quality indicators such as FCF conversion and accrual ratio
• Notes on US and Hong Kong reporting calendars

What the skill does not include

The skill contains no live market data and no data source. The agent must still obtain estimates, prices and financial statements from somewhere else. If you do not supply figures, the agent may invent numbers, so check every figure it quotes against a primary source such as SEC filings.

The return ranges in the file are generic. They are not validated for any specific stock or period, so they are not a trading system.

Install the earnings-revision skill

Download the ZIP file from this page and unzip it. You can install the skill in one of two places, depending on whether you want it for all your work or one project.

For personal use across projects, place the unzipped folder in ~/.claude/skills/. For a single project, place it in the project's .claude/skills/ folder. The skill is a single file, SKILL.md, inside the folder.

In the Claude apps you can instead upload the skill under Settings, then Capabilities, then Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

How to trigger the skill with natural requests

You do not need a special command. Ask about a company's earnings in plain language that touches the skill's topics: estimate revisions, surprises, guidance, or post-earnings drift. The agent decides whether the skill applies based on the skill's description.

Name the ticker and the market where possible, and include the figures you have. The skill's rules work on actual EPS, consensus EPS and prior guidance, so giving those numbers lets the agent apply them.

• Is consensus EPS for AAPL being revised up or down over the last 30 to 90 days? Use the earnings-revision framework.
• Here is the actual EPS and consensus for this quarter. How big is the surprise and what drift would the PEAD rules suggest?
• The company changed its full-year outlook to the upper half of the range. Score this guidance change.
• Run an earnings-quality check on this 10-Q using FCF conversion and accrual ratio.
• Give me a pre-earnings checklist for a Hong Kong stock that reports in its interim season.

Worked example: scoring a surprise

The skill defines standardized unexpected earnings as the gap between actual and consensus EPS, divided by the standard deviation of forecast errors. Values above +2 count as a large positive surprise, and values between -0.5 and +0.5 count as in-line.

If you give the agent actual EPS, consensus EPS and a historical error figure, it can compute SUE and place the result in the skill's bands. Check that the inputs are correct before you rely on the result.

Safety notes: scripts and network

The skill runs no scripts and uses no network access, according to its listing. It has no required dependencies. It is a text file the agent reads, so installing it does not run code on your machine.

Its formulas are guidelines and are not checked against any stock or period. The skill also does not tell the agent where to get data, so verify inputs yourself. Treat the output as a structured starting point for research, not as a buy or sell instruction.

• Runs scripts: no
• Uses network: no
• Requires: nothing listed
• Licence: MIT

Tips for getting useful results

Give the agent the numbers rather than asking it to find them, because the skill cannot supply them and invented figures are a known risk. Ask for the framework step by step, such as the pre-earnings checklist first and the post-earnings checklist after the report.

Use the skill for research structure and for explaining terms. Use an independent review before any decision that involves money.

• Supply actual EPS, consensus, prior guidance and cash-flow figures in your request
• Ask for one checklist at a time to keep the output organised
• Ask the agent to show its inputs so you can check them
• Remember the framework is written for US and Hong Kong equities; thresholds are not calibrated for other exchanges