Reproduce a known chat state in the running desktop on demand (model provider errors and retries, Code Mode and Den script failures, parallel tool steps, app creation, sequential commands) and check how it renders. Use when fixing or designing chat error, retry, tool-step or app-card UI.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Each state below fires in a new session of the running desktop window through drive-desktop-cdp. Provider states also need fake-model-faults set up.
export CDP_URL=http://127.0.0.1:<cdp port>
R=.opencode/skills/reproduce-chat-states/scripts/repro.mjs
C=.opencode/skills/drive-desktop-cdp/scripts/cdp.mjs
node $R provider-refused && node $C wait-idle && node $C shot /tmp/refused.png
Run node $R with no argument for the list. Pass a second argument (or set GATEWAY_MODEL) to change the real model used for the non-fault states.
| State | Fires | Check |
|---|---|---|
provider-refused, provider-dns, provider-reset | engine transport error, retried, then final | one error per turn, plain title, raw text only behind the details icon |
provider-retrying | same as reset; watch the first 30s | "Reconnecting to the model" with attempt progress, no raw text |
provider-midstream, provider-stall | cut or stalled reply | interrupted reply keeps partial text |
http-429, http-5xx, http-401, http-402 | HTTP error bodies | the matching plain-language title and next action |
codemode-wrong-tool | tools.search(...) in Code Mode | a recovered step must not read as a hard failure |
den-script-rejected | Den script with .then | Den's real reason in details, not a generic placeholder |
den-script-service-failure | provider op fails inside a Den script | "Script on OpenWork Cloud failed", reason behind the icon |
den-script-parallel | 20 Slack calls in one script | one readable group, collapsed until opened |
parallel-searches | 10 Exa searches plus Slack | clean service names and logos, no routing ids |
app-create | MCP app build with retries | one card per app that advances through its steps |
sequential-commands | 5 separate bash calls | the chat does not jump (see measure-chat-flicker) |
Stored v2 history shows exactly what the engine recorded (useful when a step looks wrong):
sqlite3 ~/.config/openwork/opencode-v2/state/opencode.db \
"select type, substr(data,1,400) from session_message where session_id='<ses_…>' order by seq"
For a step inside Code Mode, check state.metadata.toolCalls and state.metadata.openworkMcpResults on the execute part.
まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Add a new feature, put work behind a feature flag, roll something out (per organization, for everyone, cloud or self-hosted), revert or kill a feature, or make a big change to existing behavior such as a new engine. Use BEFORE writing the feature code, and when finishing or removing a rollout.
日本語の概要は準備中です。原文の説明を表示しています。
Local OpenWork Electron browser automation with CDP. Use when driving a local Electron dev app, browser_list, browser_snapshot, browser_eval, composer automation, or local UI smoke tests.
日本語の概要は準備中です。原文の説明を表示しています。
Flag customer, prospect, partner, or outside-person identities in the diff. Reported in the Warden security summary.
日本語の概要は準備中です。原文の説明を表示しています。
Connect the OpenWork MCP Gateway (https://api.openworklabs.com/mcp/agent) to Claude Code, Codex, Gemini CLI, Cursor, VS Code, Claude Desktop, or ChatGPT so the agent can use the user's OpenWork organization skills, plugins, and connections.
日本語の概要は準備中です。原文の説明を表示しています。
Screen a pull request from an outside contributor for hidden, obfuscated, or supply-chain-risky changes before any of its code runs.
日本語の概要は準備中です。原文の説明を表示しています。
Create an OpenCode plugin for OpenWork. Scaffolds the plugin file with the correct API shape, tool definitions, and hook registration. Use when the user asks to 'create a plugin', 'write a plugin', or 'make a plugin that does X'.
日本語の概要は準備中です。原文の説明を表示しています。