parallel-agents

name	parallel-agents
description	Use only when the user explicitly wants delegation, subagents, or parallel task decomposition across independent workstreams.

Native Parallel Agents

Orchestration through Codex subagents and bounded parallel work

Overview

This skill enables coordinating multiple specialized subagents through Codex-native delegation. Unlike external orchestration scripts, this approach keeps planning, delegation, and synthesis inside the same Codex workflow.

When to Use Orchestration

✅ Good for:

• Complex tasks requiring multiple expertise domains
• Code analysis from security, performance, and quality perspectives
• Comprehensive reviews (architecture + security + testing)
• Feature implementation needing backend + frontend + database work

❌ Not for:

• Simple, single-domain tasks
• Quick fixes or small changes
• Tasks where one agent suffices

---

Native Agent Invocation

Single Agent

Use the security_auditor subagent to review authentication

Sequential Chain

First, use the explorer subagent to discover project structure.
Then, use the backend_specialist subagent to review API endpoints.
Finally, use the test_writer subagent to identify test gaps.

With Context Passing

Use the frontend_specialist subagent to analyze React components.
Based on those findings, have the test_writer subagent generate component tests.

Resume Previous Work

Resume agent [agentId] and continue with additional requirements.

---

Orchestration Patterns

Pattern 1: Comprehensive Analysis

Agents: explorer → [domain-subagents] → synthesis

1. explorer: Map codebase structure
2. security_auditor: Security posture
3. backend_specialist: API quality
4. frontend_specialist: UI/UX patterns
5. test_writer: Test coverage
6. Synthesize all findings

Pattern 2: Feature Review

Agents: affected-domain-subagents → test_writer

1. Identify affected domains (backend? frontend? both?)
2. Invoke relevant domain subagents
3. test_writer verifies changes
4. Synthesize recommendations

Pattern 3: Security Audit

Agents: security_auditor → reviewer or backend_specialist → synthesis

1. security_auditor: Configuration and code review
2. reviewer or backend_specialist: Confirm impact in changed code paths
3. Synthesize with prioritized remediation

---

Available Subagents

Subagent	Expertise	Trigger Phrases
planner	Planning	"plan", "roadmap", "milestones"
explorer	Discovery	"explore", "map", "structure"
security_auditor	Security	"security", "auth", "vulnerabilities"
backend_specialist	Backend	"API", "server", "Node.js", "Express"
frontend_specialist	Frontend	"React", "UI", "components", "Next.js"
test_writer	Testing	"tests", "coverage", "TDD"
devops_engineer	DevOps	"deploy", "CI/CD", "infrastructure"
database_architect	Database	"schema", "Prisma", "migrations"
mobile_developer	Mobile	"React Native", "Flutter", "mobile"
debugger	Debugging	"bug", "error", "not working"
documentation_writer	Documentation	"write docs", "create README", "generate API docs"
performance_optimizer	Performance	"slow", "optimize", "profiling"
seo_specialist	SEO	"SEO", "meta tags", "search ranking"
reviewer	Review	"review", "regression", "correctness"
docs_researcher	External docs	"docs", "API behavior", "framework docs"

---

Built-in Generalists

These can work alongside specialized subagents:

Agent	Model	Purpose
Explore-like read-only agent	lightweight model	Fast read-only repository search
Planning-focused agent	reasoning-oriented model	Research during planning
General-purpose agent	coding model	Complex multi-step modifications

Use generalist discovery for quick searches, and specialized subagents for domain-specific analysis or implementation.

---

Synthesis Protocol

After all agents complete, synthesize:

## Orchestration Synthesis

### Task Summary
[What was accomplished]

### Agent Contributions
| Agent | Finding |
|-------|---------|
| security_auditor | Found X |
| backend_specialist | Identified Y |

### Consolidated Recommendations
1. **Critical**: [Issue from Agent A]
2. **Important**: [Issue from Agent B]
3. **Nice-to-have**: [Enhancement from Agent C]

### Action Items
- [ ] Fix critical security issue
- [ ] Refactor API endpoint
- [ ] Add missing tests

---

Best Practices

1. Available subagents - Orchestrate only the roles that materially improve the task
2. Logical order - Discovery → Analysis → Implementation → Testing
3. Share context - Pass relevant findings to subsequent agents
4. Single synthesis - One unified report, not separate outputs
5. Verify changes - Include test_writer or another validation-oriented subagent for code modifications when useful

---

Key Benefits

• ✅ Single session - All agents share context
• ✅ Codex-controlled - Orchestration stays inside the same Codex workflow
• ✅ Native integration - Works with repository skills, workflows, and subagents
• ✅ Resume support - Can continue previous agent work
• ✅ Context passing - Findings flow between agents