Kilo Code v7.3.46 adds model-specific reasoning overrides for subagents
Kilo Code `v7.3.46` ships six patch fixes covering model-specific reasoning overrides for task subagents, reversible reasoning on custom providers, session redo restoration, worktree-aware question routing, and a guard against skill removal recursively deleting working directories.
Score breakdown
The reasoning override support for subagents closes a configuration gap that previously prevented per-subagent model and variant customization, while the recursive-deletion guard removes a data-loss risk tied to skill removal.
- 01Model-specific reasoning overrides are now supported for task subagents, including custom subagents with their own model and variant settings (PR #11184).
- 02Reasoning can now be removed from custom provider models after it has been enabled (PR #11238).
- 03Reverted sessions are restored on the first Redo click (PR #11183).
Kilo Code `v7.3.46` is a patch release addressing six issues across subagent configuration, session management, worktree routing, and file safety.
A companion fix also makes reasoning toggleable on custom provider models — users can now remove reasoning after it has been enabled, closing a gap where the setting was effectively one-way.
On the reasoning configuration front, the release introduces model-specific reasoning overrides for task subagents, including custom subagents that carry their own model and variant settings. A companion fix also makes reasoning toggleable on custom provider models — users can now remove reasoning after it has been enabled, closing a gap where the setting was effectively one-way.
Three additional fixes address reliability and safety concerns: reverted sessions are now correctly restored on the first Redo click; question responses are routed to the specific worktree where the question originated rather than a default context; and skill removal is guarded against recursively deleting working directories. A UX improvement rounds out the release, clarifying loop permission prompts and auto-approval rules by displaying localized tool names.
Key facts
- 01Model-specific reasoning overrides are now supported for task subagents, including custom subagents with their own model and variant settings (PR #11184).
- 02Reasoning can now be removed from custom provider models after it has been enabled (PR #11238).
- 03Reverted sessions are restored on the first Redo click (PR #11183).
- 04Question responses are routed to the worktree where the question was created (PR #11169).
- 05Loop permission prompts and auto-approval rules now display localized tool names for clarity (PR #11232).
- 06Skill removal is prevented from recursively deleting working directories (PR #11240).
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 15, 2026 · 11:57 UTC. How this works →