# Coordinate Subagent Development

Complex implementation plans can lose quality when tasks share stale context. This skill guides isolated subagent execution with review checkpoints after every task.

## Install

```bash
npx skillstore add dmjgilbert/subagent-development
```

## Metadata

- Status: approved
- Slug: dmjgilbert-subagent-development
- Skillstore revision: r1
- Version status: missing
- Tree hash: 0b20802d72b5d8ee395ef98676432af928e35cbaf1ced866d78cdd33c691befb
- Author: DMJGilbert
- GitHub username: DMJGilbert
- License: MIT
- Repository: https://github.com/DMJGilbert/.config/tree/main/users/darren/config/claude/skills/subagent-development
- Ref: 72d5025b022c77f7a51bdf5c1637c689c80e89d1
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/dmjgilbert-subagent-development
- Manifest: https://skillstore.pages.dev/api/skills/dmjgilbert-subagent-development/manifest

## Capabilities

- Turns an existing implementation plan into tracked task work.
- Directs one fresh implementation subagent for each task.
- Adds code review checkpoints after every completed task.
- Defines actions for critical, high, medium, and low review findings.
- Provides a final review template before merge or pull request work.

## Use Cases

- Execute a phased implementation plan: Run each plan task in a fresh context and review work before moving forward.
- Add review gates to AI-assisted changes: Use reviewer checkpoints to catch regressions, missing tests, and requirement drift after each task.
- Resume long-running work safely: Track task status and deferred issues so work can pause without losing context.

## Prompt Templates

### Run a plan step by step

```
Use subagent-development on this plan. Create the task list, dispatch one fresh subagent per task, review after each task, and stop for critical issues.
```

### Implement with TDD gates

```
Use this skill to implement the plan with TDD. For each task, request a failing test, minimal implementation, passing tests, and review.
```

### Coordinate dependent tasks

```
Use subagent-development, but identify dependencies before dispatch. Only start the next task after critical and high review findings are resolved.
```

### Audit a completed task sequence

```
Apply the final review step to these commits. Check plan compliance, architecture, regressions, and unresolved review items before recommending merge readiness.
```

## Limitations

- Requires an existing implementation plan with mostly independent tasks.
- Does not define project architecture or product requirements.
- Assumes subagent and reviewer tools are available in the host assistant.
- Can add process overhead for small changes.

## Best Practices

- Start with a reviewed plan before using this skill.
- Keep each subagent task small enough to test and review.
- Resolve critical and high review findings before continuing.

## Anti Patterns

- Do not use it for quick edits that do not need task isolation.
- Do not run multiple implementation subagents in parallel.
- Do not skip reviewer feedback because tests passed.

## Security Audit

- Audited at: 2026-07-21T07:39:47.654\+00:00
- Summary: All 12 static detections are false positives. The nine external-command detections match Markdown code fences, and the three network-reconnaissance detections match ordinary implementation and review language. No prompt injection, data exfiltration, or unsafe operational intent was found in SKILL.md.

## Stats

- Views: 203
- Downloads: 10
- Favorites: 0
- Popularity score: 0