Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →lingtai-ai/preset-skill-custom
Official-source-led manual for the TUI `custom` template.
Passed 19 of 22 test cases with the skill, and 4 of 22 without it.
and 24 more in the full report
15 fixed4 passed either way3 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 15 test cases the skill fixed. Same model, same prompt, with and without the skill.
When configuring the official built-in OpenAI provider preset in the TUI codebase, what universal latest default model and vision binding should be set?
| Checked | Without skill | With skill |
|---|---|---|
| The response identifies gpt-4o as an official default model for built-in OpenAI provider presets. | Passed | Passed |
| The response notes that custom presets differ by having no provider-wide model promise. | Missed | Passed |
The instructions your model reads when the skill loads.
Use this child for the named built-in custom preset. customPreset in tui/internal/preset/preset.go:1532 is an OpenAI-compatible user-supplied template: empty model, LLM_API_KEY, user-supplied base_url, web_search, skills, and vision inherited from the configured endpoint. It has no universal latest model and no provider-wide model or vision promise.
There is no universal official catalog for custom. Read the configured endpoint provider's official model and capability documentation, then query that endpoint's authenticated GET <base_url>/models when it implements the OpenAI list operation. If it does not, use its own official catalog or CLI surface and record the exact served ID, route, protocol, and capability evidence. The OpenAI, Anthropic, and Gemini API references are protocol references only; none identifies an unspecified endpoint.
Start at customPreset in tui/internal/preset/preset.go. Custom has no providerModels or modelHasVision picker binding: model and base_url are free text. For capability/transport behavior, inspect isCustomOpenAI, isCustomOpenAIResponses, customResponsesThinkingOptions, responsesTransportOptions, and mandatoryCapRow in tui/internal/tui/preset_editor.go. Revise the constructor's inherited vision declaration only when the configured contract changes; do not turn an arbitrary relay into a universal native route.
For custom plus api_compat openai, wire_api responses exposes HTTP (omitted default) or WebSocket (responses_transport websocket), and the reasoning selector uses default omission or none/minimal/low/medium/high/xhigh. The editor removes stale transport/thinking fields outside that exact scope; a WebSocket rejection does not silently fall back to HTTP.
Prepare an evidence-bound manifest and explicit input, then run
test cases passed, vs 4 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.