fix: wire worker model from preferences through to lane runner

[NerfEko]

Summary

Mirrors the existing reviewer model pipeline (buildReviewerEnv → TASKPLANE_REVIEWER_MODEL → extraEnvVars → LaneRunnerConfig) for workers, so taskRunner.worker.model in preferences.json actually takes effect.

Type of Change

• fix (bug fix)

Problem

The reviewer model flows through a dedicated pipeline:

buildReviewerEnv() → TASKPLANE_REVIEWER_MODEL → extraEnvVars → LaneRunnerConfig.reviewerModel → --model flag

No equivalent exists for workers. Three specific gaps:

1. TaskRunnerConfig has reviewer?: { model, thinking, tools, … } but workers only get workerExcludeExtensions?: string[] — no model field
2. buildReviewerEnv() exists but no buildWorkerEnv()
3. LaneRunnerConfig hardcodes workerModel: "" while reviewerModel reads from env

Changes

File	Change
types.ts	Add worker?: { model, thinking, tools, excludeExtensions } to TaskRunnerConfig
config-loader.ts	Add worker field to toTaskRunnerConfig() return value
execution.ts	Add buildWorkerEnv(), add workerConfig param to executeWave(), wire into extraEnvVars, read workerModel/workerTools/workerThinking from env vars
engine.ts	Pass worker config in executeWave() calls, wire buildWorkerEnv() into executeLaneV2 retries

Key before/after

execution.ts — LaneRunnerConfig builder:

- workerModel: "",
- workerTools: "read,write,edit,bash,grep,find,ls",
- workerThinking: "",
+ workerModel: extraEnvVars?.TASKPLANE_WORKER_MODEL || "",
+ workerTools: extraEnvVars?.TASKPLANE_WORKER_TOOLS || "read,write,edit,bash,grep,find,ls",
+ workerThinking: extraEnvVars?.TASKPLANE_WORKER_THINKING || "",

engine.ts — executeWave call:

+ runnerConfig?.worker ? {
+     model: runnerConfig.worker.model || "",
+     thinking: runnerConfig.worker.thinking || "",
+     tools: runnerConfig.worker.tools || "",
+     excludeExtensions: runnerConfig.worker.excludeExtensions ?? [],
+ } : undefined,
  runnerConfig?.workerExcludeExtensions ?? [],

Validation

• npm test — config-loader: 95 pass / 0 fail, lane-runner: 47 pass / 1 fail (pre-existing, fails on clean branch too)
• Manual smoke test performed — configured taskRunner.worker.model: "openrouter/moonshotai/kimi-k2.6" in preferences.json, verified kimi-k2.6 usage appears in OpenRouter dashboard. Reviewer model continues to work as before.
• taskplane doctor — all checks passed
• All test failures are pre-existing, not caused by this change

Documentation

• Docs updated for user-facing changes
• Reference docs updated (commands/config/task format), if applicable
• CHANGELOG.md updated (for release-relevant changes)

Checklist

• Changes are scoped and focused
• No secrets/private data introduced
• No breaking changes — backward compatible (empty string = inherit, as before)
• Follows reviewer model pattern exactly

Related Issues

None (new bug discovered)

[Copilot]

Pull request overview

Fixes worker agent configuration propagation so taskRunner.worker.* in preferences.json (and project config) actually affects Runtime V2 lane execution, mirroring the existing reviewer model pipeline.

Changes:

• Extends TaskRunnerConfig to include worker model/thinking/tools (+ excludeExtensions) configuration.
• Adds buildWorkerEnv() and wires worker env vars through executeWave() into executeLaneV2() / LaneRunnerConfig.
• Updates engine wave execution and retry paths to pass worker config through to execution.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

File	Description
extensions/taskplane/types.ts	Adds TaskRunnerConfig.worker shape for worker agent configuration.
extensions/taskplane/config-loader.ts	Adapts unified config → legacy task-runner config to include worker fields.
extensions/taskplane/execution.ts	Adds buildWorkerEnv(), extends executeWave() signature, and reads worker model/tools/thinking from env when building LaneRunnerConfig.
extensions/taskplane/engine.ts	Passes worker config into executeWave() and includes worker env vars in certain retry env injections.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.