subagent-driven-development-3（子 Agent 驱动开发 3）

Subagent-Driven Development

Execute plan by dispatching fresh subagent per task, with multi-stage review after each: spec compliance review (domain-specialist), then independent test verification, then code quality review.

Core principle: Fresh subagent per task + domain-specialist reviews + independent verification = high quality, fast iteration

When to Use

digraph when_to_use {
    "Have implementation plan?" [shape=diamond];
    "Tasks mostly independent?" [shape=diamond];
    "Stay in this session?" [shape=diamond];
    "subagent-driven-development" [shape=box];
    "executing-plans" [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?" -> "Stay in this session?" [label="yes"];
    "Tasks mostly independent?" -> "Manual execution or brainstorm first" [label="no - tightly coupled"];
    "Stay in this session?" -> "subagent-driven-development" [label="yes"];
    "Stay in this session?" -> "executing-plans" [label="no - parallel session"];
}

vs. Executing Plans (parallel session):

• Same session (no context switch)
• Fresh subagent per task (no context pollution)
• Domain-specialist reviewers (not generic)
• Independent test verification (not self-reported)
• Systematic failure recovery (root cause analysis)
• Faster iteration (no human-in-loop between tasks)

The Enhanced Process

digraph process {
    rankdir=TB;

    subgraph cluster_per_task {
        label="Per Task";
        "Dispatch implementer subagent (./implementer-prompt.md)" [shape=box];
        "Implementer subagent asks questions?" [shape=diamond];
        "Answer questions, provide context" [shape=box];
        "Implementer implements, tests, commits, self-reviews" [shape=box];
        "Dispatch verification subagent (./verification-prompt.md)" [shape=box];
        "Verification subagent confirms tests pass?" [shape=diamond];
        "Implementer fixes test issues" [shape=box];
        "Detect task type, route to specialist reviewer" [shape=box];
        "Specialist spec reviewer confirms code matches spec?" [shape=diamond];
        "Implementer fixes spec gaps" [shape=box];
        "Optional: Security/Accessibility review gate" [shape=box style=dashed];
        "Dispatch code quality reviewer subagent" [shape=box];
        "Code quality reviewer approves?" [shape=diamond];
        "Implementer fixes quality issues" [shape=box];
        "Mark task complete in TodoWrite" [shape=box];
    }

    "Read plan, extract all tasks with full text, note context, create TodoWrite" [shape=box];
    "More tasks remain?" [shape=diamond];
    "Implementer fails task?" [shape=diamond];
    "Use systematic-debugging, then dispatch fix subagent (./fix-subagent-prompt.md)" [shape=box];
    "Dispatch final code reviewer subagent for entire implementation" [shape=box];
    "Use superpowers:finishing-a-development-branch" [shape=box style=filled fillcolor=lightgreen];

    "Read plan, extract all tasks with full text, note context, create TodoWrite" -> "Dispatch implementer subagent (./implementer-prompt.md)";
    "Dispatch implementer subagent (./implementer-prompt.md)" -> "Implementer subagent asks questions?";
    "Implementer subagent asks questions?" -> "Answer questions, provide context" [label="yes"];
    "Answer questions, provide context" -> "Dispatch implementer subagent (./implementer-prompt.md)";
    "Implementer subagent asks questions?" -> "Implementer implements, tests, commits, self-reviews" [label="no"];
    "Implementer implements, tests, commits, self-reviews" -> "Dispatch verification subagent (./verification-prompt.md)";
    "Dispatch verification subagent (./verification-prompt.md)" -> "Verification subagent confirms tests pass?";
    "Verification subagent confirms tests pass?" -> "Implementer fixes test issues" [label="no"];
    "Implementer fixes test issues" -> "Dispatch verification subagent (./verification-prompt.md)" [label="re-verify"];
    "Verification subagent confirms tests pass?" -> "Detect task type, route to specialist reviewer" [label="yes"];
    "Detect task type, route to specialist reviewer" -> "Specialist spec reviewer confirms code matches spec?";
    "Specialist spec reviewer confirms code matches spec?" -> "Implementer fixes spec gaps" [label="no"];
    "Implementer fixes spec gaps" -> "Detect task type, route to specialist reviewer" [label="re-review"];
    "Specialist spec reviewer confirms code matches spec?" -> "Optional: Security/Accessibility review gate" [label="yes"];
    "Optional: Security/Accessibility review gate" -> "Dispatch code quality reviewer subagent";
    "Dispatch code quality reviewer subagent" -> "Code quality reviewer approves?";
    "Code quality reviewer approves?" -> "Implementer fixes quality issues" [label="no"];
    "Implementer fixes quality issues" -> "Dispatch code quality reviewer subagent" [label="re-review"];
    "Code quality reviewer approves?" -> "Mark task complete in TodoWrite" [label="yes"];
    "Mark task complete in TodoWrite" -> "More tasks remain?";
    "More tasks remain?" -> "Implementer fails task?" [label="no, next task"];
    "Implementer fails task?" -> "Use systematic-debugging, then dispatch fix subagent (./fix-subagent-prompt.md)" [label="yes"];
    "Use systematic-debugging, then dispatch fix subagent (./fix-subagent-prompt.md)" -> "Mark task complete in TodoWrite";
    "More tasks remain?" -> "Dispatch final code reviewer subagent for entire implementation" [label="no"];
    "Dispatch final code reviewer subagent for entire implementation" -> "Use superpowers:finishing-a-development-branch";
}

Prompt Templates

Core templates:

• ./implementer-prompt.md - Dispatch implementer subagent
• ./spec-reviewer-prompt.md - Generic spec compliance reviewer (fallback)
• ./code-quality-reviewer-prompt.md - Code quality reviewer

Enhanced templates:

• ./fix-subagent-prompt.md - Fix failed implementation (NEW)
• ./verification-prompt.md - Independent test verification (NEW)

Reference:

• ./references/specialist-routing.md - Domain-specialist routing guide (NEW)

Domain-Specialist Routing

Before dispatching spec compliance reviewers, analyze task description for keywords:

See references/specialist-routing.md for detailed routing instructions and prompt templates.

Example Enhanced 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]
[Create TodoWrite with all tasks]

Task 1: Hook installation script

[Detect task type: generic utility]
[Get Task 1 text and context (already extracted)]
[Dispatch implementation subagent with full task text + context]

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

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

Implementer: "Got it. Implementing now..."
[Later] Implementer:
  - Implemented install-hook command
  - Added tests, 5/5 passing
  - Self-review: Found I missed --force flag, added it
  - Committed

[Dispatch verification subagent]
Verifier: Command: npm test -- --testPathPattern=hook
         Output: PASS 5/5 tests
         ✅ Claim confirmed

[Dispatch generic spec compliance reviewer - task type unclear]
Spec reviewer: ✅ Spec compliant - all requirements met, nothing extra

[Get git SHAs, dispatch code quality reviewer]
Code reviewer: Strengths: Good test coverage, clean. Issues: None. Approved.

[Mark Task 1 complete]

Task 2: Add user authentication API endpoint

[Detect task type: Backend - keywords: API, endpoint, auth]
[Dispatch implementation subagent]

Implementer: [No questions, proceeds]
Implementer:
  - Implemented POST /api/auth/login endpoint
  - Added JWT token generation
  - Tests: 12/12 passing
  - Committed

[Dispatch verification subagent]
Verifier: ✅ Claim confirmed - 12/12 tests passing

[Dispatch backend-development:backend-architect for spec review]
Backend specialist: ✅ Spec compliant
  Notes: Good RESTful design, appropriate status codes,
         but see code quality review for security concerns

[Optional: Add security review gate for auth task]
Security reviewer: Issues (Important):
  - Password not hashed before comparison
  - No rate limiting on login endpoint
  - Missing CORS configuration

[Implementer fixes security issues]
Implementer: Added bcrypt hashing, rate limiter middleware, CORS config

[Security reviewer re-checks]
Security reviewer: ✅ Approved

[Dispatch code quality reviewer]
Code reviewer: Strengths: Clean, well-tested. Issues: None. Approved.

[Mark Task 2 complete]

...

[After all tasks]
[Dispatch final code-reviewer]
Final reviewer: All requirements met, ready to merge

Done!

Enhanced Features

1. Independent Test Verification

Before accepting implementer's test report, dispatch verification subagent:

Why: Implementers may be mistaken, optimistic, or run tests incorrectly.

Template: Use ./verification-prompt.md

Integration: After implementer reports test results, before spec review.

2. Domain-Specialist Spec Review

Route spec compliance reviews to domain specialists based on task keywords.

Why: Generic reviewers miss domain-specific concerns (API design patterns, accessibility, infrastructure).

Template: See references/specialist-routing.md

Fallback: Use generic ./spec-reviewer-prompt.md for tasks without clear domain.

3. Optional Third Review Gate

For certain task types, add optional third review:

• Security review for auth/API/sensitive-data tasks
• Accessibility review for UI/frontend tasks

Why: Catch domain-specific security and accessibility issues before they ship.

4. Systematic Failure Recovery

When implementer fails a task:

1. Use systematic-debugging skill to investigate root cause
2. Dispatch fix subagent using ./fix-subagent-prompt.md
3. Re-run review gates

Why: "Dispatch fix subagent" was referenced but had no template. Now systematic.

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)
• Domain-specialist reviews (not generic)
• Independent verification (not self-reported)
• Systematic failure recovery

Quality gates:

• Self-review catches issues before handoff
• Independent test verification (evidence before assertions)
• Domain-specialist spec review (relevant expertise)
• Optional security/accessibility reviews
• Code quality ensures implementation is well-built

Cost:

• More subagent invocations per task
• Controller does more prep work
• But catches issues earlier (cheaper than debugging later)

Red Flags

Never:

• Skip reviews (verification, 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 (issues found = not done)
• Skip review loops (issues found = fix = re-review)
• Let implementer self-report replace independent verification
• Start code quality review before spec compliance is ✅
• Move to next task while any review has open issues
• Accept implementer's test report without verification

If subagent asks questions:

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

If reviewer finds issues:

• Implementer (same subagent) fixes them
• Reviewer reviews again
• Repeat until approved
• Don't skip the re-review

If implementer fails task:

1. Use systematic-debugging to investigate root cause
2. Dispatch fix subagent using ./fix-subagent-prompt.md
3. Don't try to fix manually (context pollution)

Integration

Required workflow skills:

• superpowers:writing-plans - Creates the plan this skill executes
• superpowers:requesting-code-review - Code review template for reviewer subagents
• superpowers:finishing-a-development-branch - Complete development after all tasks
• systematic-debugging - Root cause investigation for failed tasks (NEW)

Subagents should use:

• superpowers:test-driven-development - Subagents follow TDD for each task

Alternative workflow:

• superpowers:executing-plans - Use for parallel session instead of same-session execution

Optional integrations:

• critical-review - Pre-execution plan quality check
• accessibility-auditor - Third review gate for UI tasks
• Domain specialist agents - See references/specialist-routing.md