subagent-driven-development

name	subagent-driven-development
description	Use when executing implementation plans with independent tasks in the current session

Subagent-Driven Development

Execute plan by dispatching a fresh subagent per task, with two-stage review after each: spec compliance review first, then code quality review.

Why subagents: You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work.

Core principle: Fresh subagent per task + two-stage review (spec then quality) = high quality, fast iteration

Continuous execution: Do not pause to check in with your human partner between tasks. Execute all tasks from the plan without stopping. The only reasons to stop are: BLOCKED status you cannot resolve, ambiguity that genuinely prevents progress, or all tasks complete. "Should I continue?" prompts and progress summaries waste their time — they asked you to execute the plan, so execute it.

When to Use

digraph when_to_use {
    "Have implementation plan?" [shape=diamond];
    "Tasks mostly independent?" [shape=diamond];
    "subagent-driven-development" [shape=box];
    "Manual execution or brainstorm first" [shape=box];

    "Have implementation plan?" -> "Tasks mostly independent?" [label="yes"];
    "Have implementation plan?" -> "Manual execution or brainstorm first" [label="no"];
    "Tasks mostly independent?" -> "subagent-driven-development" [label="yes"];
    "Tasks mostly independent?" -> "Manual execution or brainstorm first" [label="no - tightly coupled"];
}

Why this approach:

• Fresh subagent per task (no context pollution)
• Two-stage review after each task: spec compliance first, then code quality
• Faster iteration (no human-in-loop between tasks)
• Orchestrator's own context preserved for coordination work

The Process

digraph process {
    rankdir=TB;

    subgraph cluster_per_task {
        label="Per Task";
        "Dispatch implementer subagent (./prompts/implementer-prompt.md)" [shape=box];
        "Implementer subagent asks questions?" [shape=diamond];
        "Answer questions, provide context" [shape=box];
        "Implementer subagent implements, tests, commits, self-reviews" [shape=box];
        "Dispatch spec reviewer subagent (./prompts/spec-reviewer-prompt.md)" [shape=box];
        "Spec reviewer subagent confirms code matches spec?" [shape=diamond];
        "Implementer subagent fixes spec gaps" [shape=box];
        "Dispatch code quality reviewer subagent (./prompts/code-quality-reviewer-prompt.md)" [shape=box];
        "Code quality reviewer subagent approves?" [shape=diamond];
        "Implementer subagent fixes quality issues" [shape=box];
        "Mark task complete in update_plan" [shape=box];
    }

    "Read plan, extract all tasks with full text, note context, create update_plan" [shape=box];
    "More tasks remain?" [shape=diamond];
    "Dispatch final code reviewer subagent for entire implementation" [shape=box];
    "Use finishing-a-development-branch" [shape=box style=filled fillcolor=lightgreen];

    "Read plan, extract all tasks with full text, note context, create update_plan" -> "Dispatch implementer subagent (./prompts/implementer-prompt.md)";
    "Dispatch implementer subagent (./prompts/implementer-prompt.md)" -> "Implementer subagent asks questions?";
    "Implementer subagent asks questions?" -> "Answer questions, provide context" [label="yes"];
    "Answer questions, provide context" -> "Dispatch implementer subagent (./prompts/implementer-prompt.md)";
    "Implementer subagent asks questions?" -> "Implementer subagent implements, tests, commits, self-reviews" [label="no"];
    "Implementer subagent implements, tests, commits, self-reviews" -> "Dispatch spec reviewer subagent (./prompts/spec-reviewer-prompt.md)";
    "Dispatch spec reviewer subagent (./prompts/spec-reviewer-prompt.md)" -> "Spec reviewer subagent confirms code matches spec?";
    "Spec reviewer subagent confirms code matches spec?" -> "Implementer subagent fixes spec gaps" [label="no"];
    "Implementer subagent fixes spec gaps" -> "Dispatch spec reviewer subagent (./prompts/spec-reviewer-prompt.md)" [label="re-review"];
    "Spec reviewer subagent confirms code matches spec?" -> "Dispatch code quality reviewer subagent (./prompts/code-quality-reviewer-prompt.md)" [label="yes"];
    "Dispatch code quality reviewer subagent (./prompts/code-quality-reviewer-prompt.md)" -> "Code quality reviewer subagent approves?";
    "Code quality reviewer subagent approves?" -> "Implementer subagent fixes quality issues" [label="no"];
    "Implementer subagent fixes quality issues" -> "Dispatch code quality reviewer subagent (./prompts/code-quality-reviewer-prompt.md)" [label="re-review"];
    "Code quality reviewer subagent approves?" -> "Mark task complete in update_plan" [label="yes"];
    "Mark task complete in update_plan" -> "More tasks remain?";
    "More tasks remain?" -> "Dispatch implementer subagent (./prompts/implementer-prompt.md)" [label="yes"];
    "More tasks remain?" -> "Dispatch final code reviewer subagent for entire implementation" [label="no"];
    "Dispatch final code reviewer subagent for entire implementation" -> "Use finishing-a-development-branch";
}

OpenClaw Mechanics

Dispatching Subagents

Use sessions_spawn to dispatch leaf agents, then call sessions_yield to end your turn and wait for results. Results arrive as the next user message.

Implementer dispatch:

sessions_spawn({
  agentId: "implementer",
  mode: "run",
  runTimeoutSeconds: 600,
  task: "<full implementer prompt — see ./prompts/implementer-prompt.md>"
})

Spec compliance reviewer dispatch:

sessions_spawn({
  agentId: "reviewer",
  mode: "run",
  runTimeoutSeconds: 600,
  task: "<spec reviewer prompt — see ./prompts/spec-reviewer-prompt.md>"
})

Code quality reviewer dispatch:

sessions_spawn({
  agentId: "reviewer",
  mode: "run",
  runTimeoutSeconds: 600,
  task: "<code quality reviewer prompt — see ./prompts/code-quality-reviewer-prompt.md>"
})

Final full-implementation reviewer dispatch:

sessions_spawn({
  agentId: "reviewer",
  mode: "run",
  runTimeoutSeconds: 600,
  task: "<full code reviewer prompt covering all tasks>"
})

Feedback Loop

Subagent sessions are ephemeral — they do not persist between dispatches. If a reviewer finds issues, spawn a new implementer subagent with the issues embedded directly in the task prompt. Do not attempt to reuse or continue an old session.

yield Pattern

After dispatching one or more subagents, call sessions_yield to hand off the turn. Results come back as the next user message. Only dispatch multiple subagents at once when they are truly independent (e.g., parallel reviewers on unrelated tasks). Never dispatch multiple implementers in parallel — they can conflict.

Model Selection

Use the least powerful model that can handle each role to conserve cost and increase speed.

Note: The model names below are examples. Substitute your own provider's models.

• <fast-model> — use your fastest/cheapest model (e.g. a small/free tier model)
• <mid-model> — use a capable mid-tier model (e.g. a mid-size reasoning model)
• <top-model> — use your best model (e.g. a large frontier model)

Mechanical implementation tasks (isolated functions, clear specs, 1-2 files): use <fast-model>. Most implementation tasks are mechanical when the plan is well-specified.

Integration and judgment tasks (multi-file coordination, pattern matching, debugging): use <mid-model>.

Architecture, design, and review tasks: use <top-model>.

Task complexity signals:

• Touches 1-2 files with a complete spec → <fast-model>
• Touches multiple files with integration concerns → <mid-model>
• Requires design judgment or broad codebase understanding → <top-model>

Handling Implementer Status

Implementer subagents report one of four statuses. Handle each appropriately:

DONE: Proceed to spec compliance review.

DONE_WITH_CONCERNS: The implementer completed the work but flagged doubts. Read the concerns before proceeding. If the concerns are about correctness or scope, address them before review. If they're observations (e.g., "this file is getting large"), note them and proceed to review.

NEEDS_CONTEXT: The implementer needs information that wasn't provided. Provide the missing context and re-dispatch.

BLOCKED: The implementer cannot complete the task. Assess the blocker:

1. If it's a context problem, provide more context and re-dispatch with the same model
2. If the task requires more reasoning, re-dispatch with a more capable model
3. If the task is too large, break it into smaller pieces
4. If the plan itself is wrong, escalate to your human partner

Never ignore an escalation or force the same model to retry without changes. If the implementer said it's stuck, something needs to change.

Prompt Templates

• ./prompts/implementer-prompt.md — Dispatch implementer subagent
• ./prompts/spec-reviewer-prompt.md — Dispatch spec compliance reviewer subagent
• ./prompts/code-quality-reviewer-prompt.md — Dispatch code quality reviewer subagent

Example Workflow

You: I'm using Subagent-Driven Development to execute this plan.

[Read plan file once: docs/plans/feature-plan.md]
[Extract all 5 tasks with full text and context]
[Call update_plan with all tasks]

Task 1: Hook installation script

[Get Task 1 text and context (already extracted)]
[Dispatch implementer subagent via sessions_spawn(agentId="implementer", ...)]
[Call sessions_yield to wait for result]

Implementer: "Before I begin - should the hook be installed at user or system level?"

You: "User level (~/.config/hooks/)"

[Dispatch implementer again with answer embedded in task prompt]
[sessions_yield]

Implementer: [Later]
  - Implemented install-hook command
  - Added tests, 5/5 passing
  - Self-review: Found I missed --force flag, added it
  - Committed
  - Status: DONE

[Dispatch spec compliance reviewer via sessions_spawn(agentId="reviewer", ...)]
[sessions_yield]
Spec reviewer: ✅ Spec compliant - all requirements met, nothing extra

[Get git SHAs, dispatch code quality reviewer via sessions_spawn(agentId="reviewer", ...)]
[sessions_yield]
Code reviewer: Strengths: Good test coverage, clean. Issues: None. Approved.

[Mark Task 1 complete in update_plan]

Task 2: Recovery modes

[Dispatch implementer subagent]
[sessions_yield]

Implementer: [No questions, proceeds]
Implementer:
  - Added verify/repair modes
  - 8/8 tests passing
  - Self-review: All good
  - Committed
  - Status: DONE

[Dispatch spec compliance reviewer]
[sessions_yield]
Spec reviewer: ❌ Issues:
  - Missing: Progress reporting (spec says "report every 100 items")
  - Extra: Added --json flag (not requested)

[Dispatch NEW implementer subagent with issues embedded in task prompt]
[sessions_yield]
Implementer: Removed --json flag, added progress reporting

[Spec reviewer reviews again (new sessions_spawn(agentId="reviewer", ...))]
[sessions_yield]
Spec reviewer: ✅ Spec compliant now

[Dispatch code quality reviewer]
[sessions_yield]
Code reviewer: Strengths: Solid. Issues (Important): Magic number (100)

[Dispatch NEW implementer subagent to fix]
[sessions_yield]
Implementer: Extracted PROGRESS_INTERVAL constant

[Code reviewer reviews again]
[sessions_yield]
Code reviewer: ✅ Approved

[Mark Task 2 complete in update_plan]

...

[After all tasks]
[Dispatch final code-reviewer via sessions_spawn(agentId="reviewer", ...)]
[sessions_yield]
Final reviewer: All requirements met, ready to merge

Done!

Advantages

vs. Manual execution:

• Subagents follow TDD naturally
• Fresh context per task (no confusion)
• Parallel-safe (subagents don't interfere)
• Subagent can ask questions (before AND during work)

vs. Executing Plans:

• Same session (no handoff)
• Continuous progress (no waiting)
• Review checkpoints automatic

Efficiency gains:

• No file reading overhead (controller provides full text)
• Controller curates exactly what context is needed
• Subagent gets complete information upfront
• Questions surfaced before work begins (not after)

Quality gates:

• Self-review catches issues before handoff
• Two-stage review: spec compliance, then code quality
• Review loops ensure fixes actually work
• Spec compliance prevents over/under-building
• Code quality ensures implementation is well-built

Cost:

• More subagent invocations (implementer + 2 reviewers per task)
• Controller does more prep work (extracting all tasks upfront)
• Review loops add iterations
• But catches issues early (cheaper than debugging later)

Red Flags

Never:

• Start implementation on main/master branch without explicit user consent
• Skip reviews (spec compliance OR code quality)
• Proceed with unfixed issues
• Dispatch multiple implementation subagents in parallel (conflicts)
• Make subagent read plan file (provide full text instead)
• Skip scene-setting context (subagent needs to understand where task fits)
• Ignore subagent questions (answer before letting them proceed)
• Accept "close enough" on spec compliance (spec reviewer found issues = not done)
• Skip review loops (reviewer found issues = implementer fixes = review again)
• Let implementer self-review replace actual review (both are needed)
• Start code quality review before spec compliance is ✅ (wrong order)
• Move to next task while either review has open issues
• Reuse an old subagent session for a retry — always spawn fresh

If subagent asks questions:

• Answer clearly and completely
• Provide additional context if needed
• Don't rush them into implementation

If reviewer finds issues:

• Spawn a NEW implementer subagent with the issues embedded in the task prompt
• Reviewer reviews again (new spawn)
• Repeat until approved
• Don't skip the re-review

If subagent fails task:

• Dispatch fix subagent with specific instructions
• Don't try to fix manually (context pollution)

Integration

Required workflow skills:

• using-git-worktrees — Ensures isolated workspace (creates one or verifies existing)
• writing-plans — Creates the plan this skill executes
• requesting-code-review — Code review template for reviewer subagents
• finishing-a-development-branch — Complete development after all tasks

Subagents should use:

• test-driven-development — Subagents follow TDD for each task