How to use the hv-analysis agent skill (横纵分析法)

What the hv-analysis skill makes the agent do

The skill is built on the horizontal-vertical analysis method (横纵分析法), which the source credits to Khazix. The vertical axis traces the subject's full history from its origin to today, covering its founders, funding, key versions, strategy shifts and controversies. The horizontal axis compares the subject with its competitors at the current point in time. The agent then combines both axes into a long written report.

The final output is a formatted PDF with a cover page, headers and page numbers. The skill also sets a writing style for the report, so the result reads as a narrative rather than a list of facts.

• Vertical axis: origin, timeline, decision logic and stages of development
• Horizontal axis: competitors, user reputation, ecosystem position and trends
• Intersection: links history to today's position and gives forecast scenarios
• Deliverable: a PDF named after the subject, produced by the bundled script

Install hv-analysis (ZIP or upload)

Download the ZIP from the page, then choose one of two routes. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. Either way the skill folder should contain SKILL.md, scripts/md_to_pdf.py and references/schema.json.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

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

Install the Python packages the PDF script needs

The PDF step runs a Python script that depends on two packages. The skill's own instructions give this install command, which includes a flag needed on some systems.

• Requires Python 3
• Packages: weasyprint and markdown

How to trigger the skill with a natural request

You do not need a special command. Describe what you want researched in plain language. The skill is meant for requests that need systematic background, not simple definitions. Trigger phrases in the source include 横纵分析, 研究一下, 帮我分析, 深度研究, 竞品分析, 帮我摸清楚 and 帮我搞懂, as well as deep research in English. Phrases like 'what is X' should not trigger it.

If you already name the subject clearly, the agent should start without asking follow-up questions. Otherwise it confirms the subject, its type, your motive and any focus area.

• Good: 用横纵分析法研究 Hermes Agent (the source's own example pattern)
• Good: 帮我做个 deep research on this company before we decide
• Good: 竞品分析 for a tool and its rivals, with the report as PDF
• Avoid: 什么是 X (a plain definition; the skill is wasteful here)

Example workflow from request to PDF

Once triggered, the agent is told to run web searches, ideally with parallel sub-agents, one for history and one for competitors. It checks sources and fills gaps before writing. It then writes the report as Markdown and converts it using the bundled script.

The conversion command in the source takes an input Markdown file, an output PDF and optional title and author arguments. The file name pattern for the report is the subject name followed by 横纵分析报告.

• Step 1: web search for history and competitors, preferring primary sources
• Step 2: write the vertical analysis, the horizontal analysis and the intersection
• Step 3: save as Markdown, then convert to PDF with the script
• Step 4: list sources with URLs and access dates at the end

What the skill needs to run

Beyond Python 3 and the two packages, the agent environment needs web search and web fetch tools. The source says the skill uses network access. An optional web-access skill can be installed and the agent will prefer it when present. For academic topics, the source suggests querying the arXiv API.

• Python 3 with weasyprint and markdown
• Web search and fetch tools in the agent environment
• Optional: web-access skill, if installed
• Optional: access to the arxiv.org API for academic topics
• The skill runs scripts and uses the network

Safety notes: scripts and network use

This skill runs a script and sends search queries to the public web and to third-party sites. Do not use it for confidential or internal research. Treat the output as a draft: the source warns that the agent may write long narrative claims from web sources that need checking.

Review the script before running it if you want to know exactly what it does. Check every key fact in the report against its cited source, especially for time-sensitive topics.

• Do not use for confidential or internal research
• Verify facts and dates in the report against sources
• Read scripts/md_to_pdf.py if you need to audit it
• Expect the agent to flag missing information rather than invent it, per the source's rules

Tips for better reports

Give the agent the subject's exact name, its type (product, company, concept or person) and, if relevant, why you are researching it now. The method depends on rich sources, so the skill tells the agent to search more rather than stop early. Expect long output: the source sets targets of 10,000 to 30,000 Chinese characters for a full report.

• Name the subject precisely and state its type
• Say what decision the report supports, to focus the analysis
• Ask for a specific focus area if you have one
• Check the sources list at the end before relying on the report