Data2Story — Agentic Data Storytelling¶

Development line: project:data2story · thread agentic-data-storytelling
Last event: 2026-06-16 · 1 dated since 2026-06-16 · Researched: 2026-09-05 · confidence: high

What it is¶

Data2Story turns a CSV, data folder, paper, or an idea in Pro into a self-contained HTML story. It generates charts and media alongside an evidence viewer.

• Seven newsroom roles structure the canonical drafting pipeline.
• Claim tracing links each assertion and visual back to data, code, or citations.
• Pro adds a 14-agent workflow and in-page recomputation.

The paper evaluated 18 articles with 53 participants. Use it when evidence traceability is a delivery requirement, not to replace editorial judgment.

Development line¶

• 2026-06-16 — Data2Story was publicly linked with its website and GitHub repository. On 2026-06-16, public links connected Data2Story to its project website and the data2story-skill GitHub repository. This sets the reference point for the agentic data-storytelling line without claiming unverified features.

What changed¶

• 2026-05-11 — Data2Story was accepted as a Spotlight paper at the ACM Conference on AI and Agentic Systems workshop.
• 2026-06-09 — The paper and repository were released. The canonical implementation established a seven-role sequential newsroom and evaluated 18 articles with 53 participants.
• 2026-06-16 — Data2Story was publicly presented as a dataset-to-verifiable-story workflow. The Inspector traces each claim and visual back to data, code, or references.
• 2026-06 — Version 0.1.0 optimized visual effects. The repository identifies data2story-pro as the richer implementation alongside the paper-reproducible seven-role skill.

How to use this¶

As of 2026-06-16, use the project website and GitHub repository as starting references for agentic data storytelling.

1. Clone or open the repository, then use the canonical data2story skill when reproducing the paper’s seven-role pipeline. — https://github.com/QinghongLin/data2story-skill
2. Provide a CSV, JSON file, data directory, paper, or—in Pro—an idea or URL; run the orchestrator against that input. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story-pro/README.md
3. Open index.html for the article and inspect claims through viewer.html in the canonical skill or the in-page inspector in Pro. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story/SKILL.md

Best practices¶

• Run the fixed stages in order and do not advance until the prior artifact is complete. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story/SKILL.md
• Archive the exact skills used for every run so a result can be traced and rechecked. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story/SKILL.md
• Use the Pro idea mode only when it can acquire and validate a real dataset; it stops rather than fabricates supporting data. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story-pro/README.md
• Treat Pro’s published duration as a planning estimate: --fast is listed at roughly 15–25 minutes and premium at roughly 2 hours 40–50 minutes. — https://raw.githubusercontent.com/QinghongLin/data2story-skill/main/skills/data2story-pro/README.md

Superseded by this¶

• Nothing marked obsolete yet.

Still unknown¶

• The repository update notes give only the month for Version 0.1.0, leaving the exact release day unverified.
• No first-party source establishes production operating cost, model-specific media-generation defaults, or a formal release changelog for Data2Story Pro.

Sources¶

Agent brief¶

• Subject: project:data2story, thread agentic-data-storytelling, 1 dated events 2026-06-16 → 2026-06-16.
• Practical note: As of 2026-06-16, practitioners can use the supplied project website and GitHub repository as the starting references when evaluating Data2Story for agentic data-storytelling work.
• Confidence: high. Dated supersedes above are the authority for what is obsolete.