Skill Store 33

Reproduce Chat States

Reproduce a known chat state in the running desktop on demand (model provider errors and retries,…

by different-ai · MIT

The skill

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.

Reproduce chat states

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.

States and what to look at

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)

Reading what happened

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.

Use this skill

Connect your agent to the Skill Store MCP server and ask for it by ref: different-ai/reproduce-chat-states-72735528958.

Download SKILL.md · Markdown version

Details

Publisher
different-ai
Category
Software Development › Testing & QA
License
MIT
Updated
2026-10-18

References

More from different-ai

Related skills

  • Chat Apps Ui Sdk Interactive UI rendered inside ChatGPT or Claude — OpenAI Apps SDK apps, MCP Apps (the @modelcont…
  • Chat Motion Overlay Generate configurable chat motion overlays from a transcript or screenshot, including plain bubbl…
  • Chat Mode Authoring Author Copilot chat modes and instructions
  • Chat Perf Run chat perf benchmarks and memory leak checks against the local dev build or any published VS C…
  • Build Chatgpt App Build, scaffold, refactor, and troubleshoot ChatGPT Apps SDK applications that combine an MCP ser…