Subagent-Driven Development - OpenAgent Version

Execute plan by dispatching fresh subagent per task, with two-stage review after each: spec compliance review first, then code quality review. With approval gates at each major step.

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

OpenAgent integration: This skill integrates approval gates from OpenAgent's safety-first philosophy. You will request approval before dispatching subagents and proceeding between tasks.

When to Use

dot

digraph when_to_use {
    "Have implementation plan?" [shape=diamond];
    "Tasks mostly independent?" [shape=diamond];
    "Stay in this session?" [shape=diamond];
    "openagent-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?" -> "openagent-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)
•Two-stage review after each task: spec compliance first, then code quality
•Faster iteration (no human-in-loop between tasks)
•Approval gates for safety

The Process with Approval Gates

dot

digraph process_openagent {
    rankdir=TB;

    subgraph cluster_startup {
        label="Startup";
        approval_read [label="⏸️ REQUEST\nAPPROVAL\nto read plan", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Read plan, extract all tasks with full text, note context, create TodoWrite" [shape=box];
    }

    subgraph cluster_per_task {
        label="Per Task";
        approval_dispatch [label="⏸️ REQUEST\nAPPROVAL\nto dispatch\nimplementer", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Dispatch implementer subagent (./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];
        approval_spec_review [label="⏸️ REQUEST\nAPPROVAL\nto dispatch\nspec reviewer", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Dispatch spec reviewer subagent (./spec-reviewer-prompt.md)" [shape=box];
        "Spec reviewer subagent confirms code matches spec?" [shape=diamond];
        "Implementer subagent fixes spec gaps" [shape=box];
        approval_code_review [label="⏸️ REQUEST\nAPPROVAL\nto dispatch\ncode quality\nreviewer", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Dispatch code quality reviewer subagent (./code-quality-reviewer-prompt.md)" [shape=box];
        "Code quality reviewer subagent approves?" [shape=diamond];
        "Implementer subagent fixes quality issues" [shape=box];
        approval_next [label="⏸️ REQUEST\nAPPROVAL\nto move to\nnext task", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Mark task complete in TodoWrite" [shape=box];
    }

    subgraph cluster_completion {
        label="Completion";
        "More tasks remain?" [shape=diamond];
        approval_final_review [label="⏸️ REQUEST\nAPPROVAL\nfor final\ncode review", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Dispatch final code reviewer subagent for entire implementation" [shape=box];
        approval_finish [label="⏸️ REQUEST\nAPPROVAL\nto finish\nbranch", shape=diamond, style=filled, fillcolor="#ffffcc"];
        "Use custom/openagent-finishing-a-development-branch" [shape=box style=filled fillcolor=lightgreen];
    }

    approval_read -> "Read plan, extract all tasks with full text, note context, create TodoWrite" [label="approved"];
    "Read plan, extract all tasks with full text, note context, create TodoWrite" -> approval_dispatch;
    approval_dispatch -> "Dispatch implementer subagent (./implementer-prompt.md)" [label="approved"];
    "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 subagent implements, tests, commits, self-reviews" [label="no"];
    "Implementer subagent implements, tests, commits, self-reviews" -> approval_spec_review;
    approval_spec_review -> "Dispatch spec reviewer subagent (./spec-reviewer-prompt.md)" [label="approved"];
    "Dispatch spec reviewer subagent (./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 (./spec-reviewer-prompt.md)" [label="re-review"];
    "Spec reviewer subagent confirms code matches spec?" -> approval_code_review [label="yes"];
    approval_code_review -> "Dispatch code quality reviewer subagent (./code-quality-reviewer-prompt.md)" [label="approved"];
    "Dispatch code quality reviewer subagent (./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 (./code-quality-reviewer-prompt.md)" [label="re-review"];
    "Code quality reviewer subagent approves?" -> approval_next [label="yes"];
    approval_next -> "Mark task complete in TodoWrite" [label="approved"];
    "Mark task complete in TodoWrite" -> "More tasks remain?";
    "More tasks remain?" -> approval_dispatch [label="yes"];
    "More tasks remain?" -> approval_final_review [label="no"];
    approval_final_review -> "Dispatch final code reviewer subagent for entire implementation" [label="approved"];
    "Dispatch final code reviewer subagent for entire implementation" -> approval_finish;
    approval_finish -> "Use custom/openagent-finishing-a-development-branch" [label="approved"];
}

Prompt Templates

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

Example Workflow with Approval Gates

code

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

⏸️ REQUEST APPROVAL: "May I read the plan and create TodoWrite tasks?"

[After approval]
[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

⏸️ REQUEST APPROVAL: "May I dispatch implementer subagent for Task 1: Hook installation script?"

[After approval]
[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/openagent/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

⏸️ REQUEST APPROVAL: "May I dispatch spec compliance reviewer for Task 1?"

[After approval]
[Dispatch spec compliance reviewer]
Spec reviewer: ✅ Spec compliant - all requirements met, nothing extra

⏸️ REQUEST APPROVAL: "May I dispatch code quality reviewer for Task 1?"

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

⏸️ REQUEST APPROVAL: "May I move to Task 2? (Task 1 complete)"

[After approval]
[Mark Task 1 complete]

Task 2: Recovery modes

⏸️ REQUEST APPROVAL: "May I dispatch implementer subagent for Task 2: Recovery modes?"

[After approval]
[Get Task 2 text and context (already extracted)]
[Dispatch implementation subagent with full task text + context]

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

⏸️ REQUEST APPROVAL: "May I dispatch spec compliance reviewer for Task 2?"

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

[Implementer fixes issues]
Implementer: Removed --json flag, added progress reporting

[Spec reviewer reviews again]
Spec reviewer: ✅ Spec compliant now

⏸️ REQUEST APPROVAL: "May I dispatch code quality reviewer for Task 2?"

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

[Implementer fixes]
Implementer: Extracted PROGRESS_INTERVAL constant

[Code reviewer reviews again]
Code reviewer: ✅ Approved

⏸️ REQUEST APPROVAL: "May I move to Task 3? (Task 2 complete)"

[After approval]
[Mark Task 2 complete]

...

[After all tasks]
⏸️ REQUEST APPROVAL: "May I dispatch final code reviewer for entire implementation?"

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

⏸️ REQUEST APPROVAL: "May I use finishing-a-development-branch to complete this work?"

[After approval]
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)
•Approval gates prevent runaway automation

vs. Executing Plans:

•Same session (no handoff)
•Continuous progress (no waiting)
•Review checkpoints automatic
•User awareness at every major step

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
•Approval gates ensure user control

Cost:

•More subagent invocations (implementer + 2 reviewers per task)
•Controller does more prep work (extracting all tasks upfront)
•Review loops add iterations
•Approval gates add human interaction
•But catches issues early (cheaper than debugging later)
•User maintains control throughout process

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
•Skip approval gates to "save time" (defeats OpenAgent safety)
•Auto-approve on behalf of user (always wait for explicit approval)

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 subagent fails task:

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

OpenAgent Approval Gate Summary

Approval points in workflow:

•⏸️ Before reading plan and creating TodoWrite
•⏸️ Before dispatching each implementer subagent
•⏸️ Before dispatching spec reviewer subagent (per task)
•⏸️ Before dispatching code quality reviewer (per task)
•⏸️ Before moving to next task
•⏸️ Before final code review
•⏸️ Before using finishing-a-development-branch

Why approval gates matter:

•User maintains control over automation
•Prevents runaway subagent dispatching
•Creates natural checkpoints for oversight
•Aligns with OpenAgent's safety-first philosophy
•User can pause/redirect at any point
•Transparency into what's happening when

Integration

Required workflow skills:

•custom/openagent-using-git-worktrees - REQUIRED: Set up isolated workspace before starting
•custom/openagent-writing-plans - Creates the plan this skill executes
•custom/openagent-requesting-code-review - Code review template for reviewer subagents
•custom/openagent-finishing-a-development-branch - Complete development after all tasks

Subagents should use:

•custom/openagent-test-driven-development - Subagents follow TDD with approval gates for each task

Alternative workflow:

•custom/openagent-executing-plans - Use for parallel session instead of same-session execution

Common Rationalizations

Excuse	Reality
"Approval gates slow me down"	Gates prevent expensive mistakes. Faster than debugging.
"User already approved the plan"	Plan approval ≠ execution approval. User needs visibility.
"Subagents are safe, don't need gates"	Automation without oversight = risk. Gates provide control.
"I'll just do all tasks then ask"	User wants to see progress. Gates enable intervention.
"Asking every time is annoying"	Transparency > convenience. User appreciates control.
"Gates are unnecessary ceremony"	Gates = checkpoints. User can redirect if needed.

All of these mean: Request approval. No exceptions.

Red Flags - STOP and Request Approval

•Dispatching subagent without approval
•Moving to next task without approval
•"User won't mind if I just..."
•"I'll batch approvals to save time"
•"This is too small to need approval"
•"Already got plan approval, that's enough"
•Assuming silence = approval
•Rationalizing that gates are optional

All of these mean: Stop. Request approval first.

Final Rule

code

Subagent dispatched → user approved first
Task transition → user approved first
Otherwise → not OpenAgent workflow

No exceptions without your human partner's explicit permission.