[FEATURE]:Specify sub-agent variant (thinking lvl) from script for token economy #7400

Open
opened 2026-02-16 18:07:03 -05:00 by yindo · 3 comments
Owner

Originally created by @Solido on GitHub (Jan 23, 2026).

Originally assigned to: @thdxr on GitHub.

Feature hasn't been suggested before.

  • I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

When writing script for divide and conquer feature, sub-tasks may not require the same effort than the main agent.

I would like to launch sub-agents from the same provider but with lower or higher thinking level.

Claude prompt can launch Haiku explicitly as an example.

The request is to configure them from the prompt ONLY.

Thanks!

Originally created by @Solido on GitHub (Jan 23, 2026). Originally assigned to: @thdxr on GitHub. ### Feature hasn't been suggested before. - [x] I have verified this feature I'm about to request hasn't been suggested before. ### Describe the enhancement you want to request When writing script for divide and conquer feature, sub-tasks may not require the same effort than the main agent. I would like to launch sub-agents from the same provider but with lower or higher thinking level. Claude prompt can launch Haiku explicitly as an example. The request is to configure them from the prompt ONLY. Thanks!
yindo added the discussionzen labels 2026-02-16 18:07:03 -05:00
Author
Owner

@github-actions[bot] commented on GitHub (Jan 23, 2026):

This issue might be a duplicate of existing issues. Please check:

  • #7138: Support default variant configuration per agent - Already implemented support for configuring variant per agent in AGENTS.md
  • #7713: Commands Model Variant Specification - Requests ability to specify variant for commands
  • #8123: I can't set the effort when configuring a sub-agent! - User asking how to configure thinking level for sub-agents
  • #7296: Allow configurable subagent-to-subagent task delegation with call limits - Related to sub-agent configuration

Your request about specifying thinking level from script for sub-agents appears to be covered or related to issue #7138 (variant per agent config). If you're looking to configure this in your agent definitions in AGENTS.md or opencode.jsonc, that feature may already be available.

Feel free to ignore if your use case requires something different than these existing issues.

@github-actions[bot] commented on GitHub (Jan 23, 2026): This issue might be a duplicate of existing issues. Please check: - #7138: Support default variant configuration per agent - Already implemented support for configuring variant per agent in AGENTS.md - #7713: Commands Model Variant Specification - Requests ability to specify variant for commands - #8123: I can't set the effort when configuring a sub-agent! - User asking how to configure thinking level for sub-agents - #7296: Allow configurable subagent-to-subagent task delegation with call limits - Related to sub-agent configuration Your request about specifying thinking level from script for sub-agents appears to be covered or related to issue #7138 (variant per agent config). If you're looking to configure this in your agent definitions in AGENTS.md or opencode.jsonc, that feature may already be available. Feel free to ignore if your use case requires something different than these existing issues.
Author
Owner

@mugnimaestra commented on GitHub (Feb 7, 2026):

Heads up — PR #12567 implements this. It reads the parent session's variant and forwards it to the child session in task.ts, so subagents now inherit the thinking level automatically. The inheritance only activates when the subagent doesn't have its own variant configured, preserving explicit per-agent control.

@mugnimaestra commented on GitHub (Feb 7, 2026): Heads up — PR #12567 implements this. It reads the parent session's `variant` and forwards it to the child session in `task.ts`, so subagents now inherit the thinking level automatically. The inheritance only activates when the subagent doesn't have its own variant configured, preserving explicit per-agent control.
Author
Owner

@Solido commented on GitHub (Feb 7, 2026):

@mugnimaestra +1 for you PR. This is really a core feature when swarming. To my surprise while central in Claude philosophy it's not much present in OpenCode community. Last demo from Claude rely on this same core when lauching judge, workers, coordinators.

Thanks!!

@Solido commented on GitHub (Feb 7, 2026): @mugnimaestra +1 for you PR. This is really a core feature when swarming. To my surprise while central in Claude philosophy it's not much present in OpenCode community. Last demo from Claude rely on this same core when lauching judge, workers, coordinators. Thanks!!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#7400