Describe the feature or problem you'd like to solve
Today, model selection in Copilot CLI is a manual step (/model). There's no way for the AI itself — or a skill/instruction — to switch models automatically based on task complexity. This matters for enterprise teams where: - Users default to expensive premium models (Opus, GPT-5.4) even for simple tasks like Q&A or formatting - There's no guardrail or nudge to right-size model usage to cost
Proposed solution
Request: Expose a way to switch the active model programmatically — either via a slash command the AI can invoke (e.g., /model triggered mid-session), a --model startup flag, or a skill API hook. This would unlock an "Auto" mode
(similar to Copilot 365 / ChatGPT Auto) where the AI classifies task complexity and routes to the appropriate model automatically — reducing overhead for users and cost for the organisation.
Workaround today: I've built an auto-model skill that classifies tasks and recommends a switch, but it can't action it — the user still has to manually confirm. Closing that last step would make this feature complete.
Example prompts or workflows
No response
Additional context
No response
Describe the feature or problem you'd like to solve
Today, model selection in Copilot CLI is a manual step (/model). There's no way for the AI itself — or a skill/instruction — to switch models automatically based on task complexity. This matters for enterprise teams where: - Users default to expensive premium models (Opus, GPT-5.4) even for simple tasks like Q&A or formatting - There's no guardrail or nudge to right-size model usage to cost
Proposed solution
Request: Expose a way to switch the active model programmatically — either via a slash command the AI can invoke (e.g., /model triggered mid-session), a --model startup flag, or a skill API hook. This would unlock an "Auto" mode
(similar to Copilot 365 / ChatGPT Auto) where the AI classifies task complexity and routes to the appropriate model automatically — reducing overhead for users and cost for the organisation.
Workaround today: I've built an auto-model skill that classifies tasks and recommends a switch, but it can't action it — the user still has to manually confirm. Closing that last step would make this feature complete.
Example prompts or workflows
No response
Additional context
No response