Share a Claude Code conversation as a public preview page
Install the optional share-claude-code-chat skill, render local Claude Code JSONL conversations as high-fidelity HTML pages, and deploy them to PreviewShip as public URLs.
Answer First
The share-claude-code-chat skill is a companion workflow for Claude Code conversations. It reads local Claude Code JSONL history, reconstructs visible user and assistant messages, keeps tool activity in readable collapsible timelines, hides reasoning text and unsafe raw tool outputs, then deploys the generated HTML with PreviewShip so teams can share research trails, implementation records, and review context.
Deploy build artifacts, not source-code zips.
If you upload a zip, build the project first and zip the static output folder such as dist, build, out, or public with index.html and assets. Do not zip raw React/Vue/Next source folders with package.json and node_modules. Single .html files are supported directly by console upload, pasted HTML, CLI, MCP, and the VS Code/Cursor extension.
Key takeaways
Recommended workflow
Install globally with `npx skills add blockdancez/PreviewShip --skill share-claude-code-chat -a codex -g --yes`, or install locally from a project root without `-g`.
Make sure PreviewShip CLI is authenticated with `npx previewship login`, or set `PREVIEWSHIP_API_KEY` for the environment.
In Codex, run `$share-claude-code-chat 分享 Claude Code 对话` and choose the current or target Claude Code conversation.
Review the generated HTML page, especially if the Claude Code session touched private repos, customer context, screenshots, or credentials.
Share the PreviewShip URL returned by the skill.
Comparison snapshot
Why Claude Code needs its own sharing skill
Claude Code stores conversation history, message types, tool calls, and intermediate events differently from Codex. Reusing a Codex-specific parser would either miss useful activity or put tool timelines in the wrong place.
The Claude Code skill treats the conversation as a separate source format. It reconstructs user and assistant messages, attaches tool activity to the surrounding assistant turn, and renders the result as a self-contained HTML page that is easier to review than screenshots or pasted terminal output.
What the shared page includes
The generated page focuses on visible collaboration: user requests, assistant responses, safe status summaries, and tool activity that explains what the agent did.
It is designed to keep the page readable even when a Claude Code session includes many web fetches, shell steps, file reads, or intermediate progress events.
Privacy and review expectations
The renderer is intentionally conservative: hidden reasoning text is not published, raw tool outputs are not dumped into the page, and secret-looking values are redacted.
Because Claude Code conversations often include private repository paths, URLs, screenshots, and operational context, users should still inspect the generated preview before making it public.
FAQ
Is share-claude-code-chat the same as share-codex-chat?
No. They share the same PreviewShip publishing goal, but Claude Code and Codex store different local transcript formats. The Claude Code skill has its own parser and tool-timeline handling.
Does the skill publish Claude Code reasoning?
No. It is designed to hide reasoning text and show only safe, visible conversation content plus summarized tool activity.
Can it show tool calls without overwhelming the page?
Yes. Tool activity is grouped into collapsible timelines so readers can expand the process when they need it and keep the main answer readable.
Does it require a PreviewShip API key?
Yes for deployment. If the CLI is not authenticated, run `npx previewship login` or configure `PREVIEWSHIP_API_KEY` before sharing.