feat(agent): add orchestrate subagent for task decomposition

[kuro-toji]

Issue for this PR

Closes #24447 (partially - addresses need for better task decomposition)

Type of change

• New feature

What does this PR do?

Add a new orchestrate subagent that breaks down complex multi-step tasks into parallel subtasks executed by specialized agents.

Problem: Large tasks require manual decomposition and coordination of multiple Task calls.

Solution: Automate task breakdown with an orchestrator that:

1. Analyzes complex tasks
2. Creates focused subtasks (2-5 items)
3. Executes subtasks in parallel via Task tool
4. Synthesizes results

Files Changed

File	Change
packages/opencode/src/agent/prompt/orchestrate.txt	New orchestrator prompt
packages/opencode/src/agent/agent.ts	Register orchestrate agent
packages/opencode/src/tool/task.txt	Document orchestrate usage

How did you verify your code works?

• Code follows existing agent patterns (explore, general, plan)
• Import added to agent.ts
• Orchestrator has appropriate permissions (task, todowrite, question)
• Task tool docs updated with orchestrate example

Checklist

• I have not included unrelated changes in this PR
• Follows existing agent registration patterns

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Based on my search results, I found one potentially related PR:

Related PR:

• PR feat(tool): add experimental team tool for parallel subagents #20152: "feat(tool): add experimental team tool for parallel subagents"
  • feat(tool): add experimental team tool for parallel subagents #20152
  • This appears related as it deals with parallel execution of subagents, which overlaps with the orchestrate agent's goal of executing subtasks in parallel.

The other results (PRs #7156, #17210, #18528) are about general subagent features but less directly related to the task decomposition and parallel execution pattern introduced in PR #24450.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[kuro-toji]

Thanks for the context! I see there's related work in PR #20152 (team tool for parallel subagents).

My orchestrate agent approach is different - it's a prompt-based subagent that does task decomposition, rather than a new tool.

Should I:

1. Close this PR and let feat(tool): add experimental team tool for parallel subagents #20152 handle parallel execution
2. Pivot to support/complement the team tool with the orchestrate prompt approach
3. Keep as-is for prompt-based task decomposition (which could work alongside the team tool)

Let me know which direction makes sense!