A structured framework for reviewing dashboards, reports, and data visualizations with the depth of a senior data analyst.

Reviews are calibrated to the dashboard's actual audience, such as executives, department heads, operations managers, analysts, frontline workers, or external stakeholders. The skill does not assume every dashboard is built for a CEO.

When you share a dashboard as a PBIX file, Tableau file, PDF, spreadsheet, image, screenshot, or text description, the framework produces a structured review covering:

• 6 scored dimensions: business question, domain knowledge, audience fit, data story, visual design, and actionability
• Internal consistency checks for arithmetic, totals, units, and deviation signs
• Time horizon checks for whether the data window supports the claims
• 5+ analytical challenge questions framed for the actual audience
• 3+ missing analyses a domain expert would run next
• Top 3 prioritized recommendations
• One specific thing done well

Recommended free path:

1. If you use ChatGPT often, create a ChatGPT Project and add CHATGPT_PROJECT_INSTRUCTIONS.md as the project instructions.
2. If you only need a quick review, use the Live Prompt Builder and paste the generated prompt into your AI chat.
3. If you use Claude and have Skills enabled, install dashboard-review.skill.
4. If you use Copilot, Gemini, Codex, or another assistant, paste SKILL.md as the instruction prompt.

This repo does not require a paid OpenAI API key, Anthropic API key, Cloudflare Worker, or Custom GPT builder subscription.

Use this when you want a reusable ChatGPT workspace for dashboard reviews.

Use CHATGPT_PROJECT_INSTRUCTIONS.md for ChatGPT Projects. It is a compact v3 version of the skill, kept under 7,500 characters so it fits ChatGPT's project instruction limit.

1. Open ChatGPT and sign in.
2. Create a new Project.
3. Name the project:

Dashboard Review

4. Inside the project, click the three-dot menu (...) in the top-right corner.
5. Open the project instructions or settings from that menu.
6. Copy the full contents of CHATGPT_PROJECT_INSTRUCTIONS.md from this repo.
7. Paste it into the Project instructions.
8. Save the project.
9. Start a new chat inside that project.
10. Upload a dashboard file, screenshot, PDF, or describe the dashboard.
11. Ask:

Review this dashboard.

Use this option when:

• You want ChatGPT to remember the review framework across multiple dashboard review chats.
• You review dashboards often.
• You do not want to create or pay for a Custom GPT.
• You want the cleanest free ChatGPT workflow.

Note: ChatGPT does not currently support custom user slash commands like Claude's /dashboard-review. A ChatGPT Project is the closest free ChatGPT-native replacement.

The GitHub Pages app is a prompt builder. It does not call Anthropic, OpenAI, or any other model API.

How it works:

1. Add dashboard context and optionally upload a PBIX, Tableau workbook, PDF, spreadsheet, screenshot, or other dashboard file.
2. The app builds a structured review prompt in your browser.
3. Copy the prompt into ChatGPT, Microsoft Copilot, Claude, Gemini, Codex, or another AI assistant.
4. If you uploaded a file, attach the same file in that AI chat.

This keeps the public app free to host and free to use. No visitor data or files are sent anywhere by the page.

Use this option when:

• You want the fastest no-setup workflow.
• You are sharing the tool with people who may use different AI assistants.
• You want a structured prompt that works with PBIX files, Tableau workbooks, PDFs, spreadsheets, screenshots, and other dashboard files in ChatGPT, Microsoft Copilot, Claude, Gemini, Codex, or another AI tool.
• You do not want users to install anything.

Use SKILL.md when the AI tool lets you paste custom instructions, system prompts, or reusable context.

1. Open SKILL.md.
2. Copy the full file.
3. Paste it into the AI tool's instructions or into the start of a new chat.
4. Add:

Use these instructions as your dashboard review framework.

5. Upload a dashboard file or describe the dashboard.
6. Ask for a review.

1. Download dashboard-review.skill from the repo or Releases page.
2. In Claude.ai, go to Settings -> Connectors -> Customize -> Skills -> Plus Sign (+) -> Upload Skill.
3. Upload the .skill file.
4. In Claude, type:

/dashboard-review

5. Share a dashboard file, screenshot, or description and ask Claude to review it.

To confirm it is working, start a new Claude conversation and ask:

is the dashboard review skill active?

Once installed or pasted into your AI tool, ask things like:

Review my dashboard
Critique this Power BI report
Give feedback on this Tableau viz
What's wrong with my dashboard?
Score my dashboard
Help me improve this report

Or upload a PBIX, Tableau file, spreadsheet, dashboard image/PDF, or any file your AI tool can inspect and ask for a review.

The skill first determines who the dashboard was built for, then applies all scoring criteria through that lens.

1. Internal consistency: add up segment totals and compare them with the stated overall.
2. Time horizon adequacy: check whether the date range is long enough for the claims.
3. Deviation sign consistency: make sure positive and negative values mean the same thing everywhere.

dashboard-review-skill/
|-- README.md                 # Project overview and setup notes
|-- SKILL.md                  # Full portable skill instructions for Copilot, Codex, Claude, and other AI tools
|-- CHATGPT_PROJECT_INSTRUCTIONS.md # Compact ChatGPT Project version under 7,500 characters
|-- dashboard-review.skill    # Packaged Claude skill archive
|-- agents/
|   `-- openai.yaml           # ChatGPT/Codex skill metadata
|-- docs/
|   `-- index.html            # Free GitHub Pages prompt builder
`-- .gitignore                # Local tooling ignores

• Refresh the packaged dashboard-review.skill after source updates.
• Add example review prompts and outputs.
• Add screenshots for the ChatGPT Project setup flow.
• Add a short guide for installing SKILL.md as a local Codex skill.

MIT. Use freely, adapt as needed, attribution appreciated.

Built by David Abu and shared with the Data Community Africa community.