Visualization Reporting

Skill: Visualization & Reporting Workflow

Objective: Transform raw numerical data and insights into beautiful, shareable assets. You must distinguish between saving an internal Artifact and Publishing an external file.

Execution Steps:

1. Confirm Data: Ensure you have gathered the necessary metrics (e.g., a dictionary of channels and their engagement rates) or the final synthesized text.
2. Visualize: Delegate to the visualization_agent to create static charts. Pass the raw data to it. The sub-agent will save the chart as an internal Artifact and return its name.
3. Retrieve Artifact: Use the load_artifacts tool to load the raw bytes of the image artifact from the session memory.
4. Publish Dependencies to GCS: If your final HTML report will display the image, you MUST publish the image bytes FIRST using publish_file(content, filename, "image/png"). This returns a public https://storage.googleapis.com/... URL.
5. Construct HTML: Create a clean, well-formatted HTML string containing your analysis. Embed the charts using the public URLs obtained in step 4 (e.g., <img src="https://storage.../chart.png" />). Do NOT use local paths.
6. Final Delivery Options:
   • If the user just wants the report saved: Call render_html(html_content, "report.html") to save it as an internal artifact they can download later.
   • If the user wants a shareable link: Call publish_file(html_content, "report.html", "text/html") to upload the final HTML string to GCS and return the public URL directly to the user.

Source: google/adk-samples → python/agents/youtube-analyst/youtube_analyst/skills/visualization-reporting/SKILL.md