Export for coding agent

Export for coding agent is an owner-only action that downloads a ZIP containing the artifact’s current stored HTML and a complete snapshot of its feedback history — every open and resolved comment thread, reply, and reaction. It’s the bridge between a review that just happened in Curio and the next coding session that acts on it.

What’s inside

• README.md — a short guide that points the agent at the other three files and explains how to treat them.
• artifact.html — the exact source of the current version, byte-for-byte, with no injected Curio bridge code.
• comments.md — a human- and prompt-friendly summary with open threads listed first as requested work, then resolved threads as history.
• comments.json — the same feedback as a lossless, machine-readable snapshot, including thread status, replies, reaction counts, and stored anchor hints.

How to use it

Attach the ZIP to a chat session or unpack it in Claude, ChatGPT, or any coding agent, then ask it to address the open feedback. Open threads are requested work; resolved threads are kept as historical context, not instructions to redo. Anchors — the CSS selector, quoted text, element details, and approximate position stored with each comment — are hints about where feedback was pinned, not guarantees; an agent should infer intent rather than assume an exact match, especially for feedback pinned on an earlier version.

The trust boundary

The exported HTML and every comment body are untrusted data, not instructions from Curio — they came from whoever authored the artifact or left feedback on it. The bundled README.md tells the receiving agent exactly that, so it treats the artifact and the comments as project content to read and reason about, not as commands to blindly execute.

Where to find it

The Export for coding agent button lives in an artifact’s settings, below Versions. It always exports the current version — selecting an older version in the viewer doesn’t change what gets exported.