Guide · Agent skill
How to use the Financial Research Report Writer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The report-generate skill teaches an AI agent to write financial research reports in the layout used by securities firms and asset managers. This guide explains what the skill changes in the agent's output, how to install it, how to ask for a report in plain language, and what to check before you rely on the result. You do not need prior experience with agent skills to follow it.
Quick start
- 1Download the ZIP. Download the skill ZIP from this page. It contains SKILL.md.
- 2Unzip into the skills folder. Unzip the file so the skill folder sits in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill.
- 4Ask for a report. Use a plain request such as Analyze stock XX, How is sector XX?, or What is the market outlook?
- 5Provide your inputs. Include the figures and time period you want covered, since the skill does not supply market data.
- 6Request a PDF if needed. Ask for a PDF and check that the agent reports a successful write_file result and a download link.
- 7Review before use. Verify numbers and claims yourself, because the output is AI-generated and not investment advice.
What the skill makes the agent do
Without the skill, an agent answers a finance question in whatever format it chooses. With the skill installed, the agent follows a fixed report structure: a summary, core views, main analysis, a data appendix, risk warnings and an investment recommendation. The skill also gives it a rating scale, Markdown table and number formatting rules, and a finance terminology guide.
The skill covers five report types: deep-dive stock reports, industry research, strategy reports, backtest reports and short flash notes. Each type has a target length range and a typical trigger, such as a request to analyse a stock or a finished backtest.
- Deep-dive stock report: 3000-5000 words, company analysis, valuation and rating.
- Industry research: 2000-4000 words, industry structure, trends and recommended names.
- Strategy report: 1500-3000 words, macro view, strategy and allocation recommendation.
- Backtest report: 1000-2000 words, strategy performance, risk and improvement suggestions.
- Flash comment or note: 500-1000 words, event commentary, impact and action suggestion.
Install the skill (ZIP download or Claude apps upload)
The skill is distributed as a ZIP file from this page. The package contains one file, SKILL.md. There are two ways to add it to your setup.
For Claude Code, unzip the downloaded file so that the skill folder sits in your personal skills directory ~/.claude/skills/, or in a project's .claude/skills/ folder if you want it only for that project.
In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill there.
- Personal use across projects: place the skill folder in ~/.claude/skills/.
- Single project only: place the skill folder in that project's .claude/skills/ folder.
- Claude apps: Settings → Capabilities → Skills, then upload.
How to trigger the skill with natural requests
You do not need a special command. Ask for a report in ordinary language and the agent will apply the skill when your request matches one of the report types. The source gives these example triggers, so phrasing close to them works well.
For a backtest report, ask after a backtest has finished, because the skill reads the results of that completed run. The skill does not supply market data, so include the stock, sector or strategy details you want covered in your request or make sure another tool can provide them.
- Analyze stock XX (deep-dive stock report)
- How is sector XX? (industry research)
- What is the market outlook? (strategy report)
- How should we view event XX? (flash comment or note)
- After a backtest is completed: ask for a backtest report on that run.
Understand the rating scales and formatting rules
Equity reports use four ratings: Strong Buy (expected 12-month return above 30%), Buy (15%-30%), Neutral (-5% to 15%) and Avoid (below -5%). Strategy reports use a separate scale based on Sharpe ratio and maximum drawdown, from High allocation priority down to Not recommended.
Tables put units in the header, use E for forecast years and A for actual years, and keep one decimal place for percentages. Because Markdown cannot embed images, charts are replaced by a text description and a data table.
- Percentages: X.X% (for example 12.5%).
- Multiples: X.Xx (for example 20.5x).
- Dates: YYYY-MM-DD.
- Forecast year: 2025E; actual year: 2024A.
- Wording: the terminology guide prefers terms such as pullback or valuation premium over informal words like dropped or expensive.
Save the report as a PDF
When you ask for a PDF, the skill sends the full Markdown report to a file-writing tool called write_file, with a .pdf target. It prefers artifacts/report.pdf under the active run directory. The tool renders a paginated PDF and does not need system Pango or Cairo installed.
The agent should check the returned status, path and bytes_written, and give you a download link labelled with the report filename. If rendering fails, the agent should report the tool error rather than claim a PDF exists. In the Web chat, the file downloads with authentication; IM channels attach the PDF.
What the skill needs and safety notes
The skill needs a file-writing tool called write_file in the host platform, but only for PDF output. Market data or other inputs must come from you or from other tools, because the source does not specify any data source. According to the file list, the skill runs no scripts and makes no network calls.
The report is AI-generated analysis and can contain errors. The skill's own disclaimer says it is for research reference only and is not investment advice. It does not add compliance review or sourcing checks, so do not publish its output as regulated research without human review.
- Requires: write_file for PDF output only.
- Data: supplied by you or other tools; none is specified by the skill.
- Scripts: none.
- Network: none.
- Review: check every number and claim against your own sources before use.
Tips for better reports
Give the agent concrete inputs. Naming the company, sector or strategy, the period and any figures you already have will produce a more specific report than a vague question.
Ask for the report type you need, because length, structure and rating scale depend on it. Check the data appendix and risk section first, since those are where missing or wrong numbers are easiest to spot.
- Name the company, sector or strategy and the time period.
- Paste the figures you trust into the request.
- Choose the report type deliberately; a flash note is much shorter than a deep-dive.
- Read the Data Appendix and Risk Warnings before the recommendation.
What people use it for
- Drafting a first version of a deep-dive stock report with a rating and valuation section for an analyst to revise.
- Producing an industry overview that lists structure, trends and candidate names in the standard layout.
- Writing a backtest report from a completed run, with equity curve substitute tables and risk notes.
- Helping students see how sell-side style research is organised and rated.
- Generating consistently formatted PDF reports for a team working in the Vibe-Trading platform.
Frequently asked questions
How do I install the report-generate skill?
Download the skill ZIP from this page and unzip it so the skill folder sits in ~/.claude/skills/ or in a project's .claude/skills/ folder. In the Claude apps, you can instead upload it under Settings, Capabilities, then Skills.
Is the report-generate skill free to use?
The source lists the licence as MIT. Check the licence file in the source repository for the full terms before you reuse or redistribute it.
What does the report-generate skill require?
It needs a file-writing tool called write_file in the host platform, but only if you want PDF output. It does not specify any data source, so market data must come from you or other tools. It runs no scripts and makes no network calls.
Can I rely on the reports for investment decisions?
No, not without human review. The skill produces AI-generated analysis that can contain errors, and its disclaimer states that it is not investment advice. It also adds no compliance review or sourcing checks.
Why is my PDF not created?
Check the write_file result for status, path and bytes_written. If rendering failed, the tool returns an error, and the agent should report that error rather than claim a PDF exists. A successful write confirms that the returned path exists.
How does this skill compare with other report tools?
The source does not compare it with other tools. It is best suited to a fast first draft in the standard research layout, and it is not a substitute for compliance review or verified data.
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
Use this skill if you want a fast, consistently structured first draft of a stock, industry, strategy or backtest report, or if you are learning how sell-side research is organised and rated. Pick something else if you need publication-ready research with compliance review, sourced data, or outputs you can act on without checking them yourself.
Download the skill