parallel-agents

name	parallel-agents
description	Multi-agent orchestration patterns for complex tasks requiring multiple domain expertise or comprehensive analysis.
version	1.0.0
allowed-tools	Read, Glob, Grep

Native Parallel Agents

Purpose: Enable coordinating multiple specialized agents through Antigravity's native agent system while maintaining Trust-Grade governance.

Overview

This skill enables coordinating multiple specialized agents for complex tasks. Unlike external scripts, this approach keeps all orchestration within the kit's control while applying Trust-Grade governance protocols.

---

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
• Trust-Grade verticals (Auth, Payment, etc.)

❌ Not for:

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

---

Native Agent Invocation

Single Agent

Use the security-reviewer agent to audit authentication

Sequential Chain

First, use the architect to review structure.
Then, use the code-reviewer to check implementation.
Finally, use the tdd-guide to verify test coverage.

With Context Passing

Use the planner to break down the feature.
Based on that plan, have the architect design the structure.
Then, have the tdd-guide create test specifications.

Resume Previous Work

Resume agent [agentId] and continue with additional requirements.

---

Orchestration Patterns

Pattern 1: Feature Vertical (Trust-Grade Standard)

Agents: planner → architect → [domain-agents] → tdd-guide → e2e-runner

1. planner: Break down feature into tasks
2. architect: Design component structure
3. [domain-agents]: Implement by domain
4. tdd-guide: Create/verify unit tests
5. e2e-runner: Create/verify E2E tests
6. Synthesize all deliverables

Pattern 2: Code Review Comprehensive

Agents: code-reviewer → security-reviewer → synthesis

1. code-reviewer: Quality and patterns review
2. security-reviewer: Security posture audit
3. Synthesize with prioritized findings

Pattern 3: Security Audit

Agents: security-reviewer → architect → tdd-guide

1. security-reviewer: Vulnerability assessment
2. architect: Secure architecture patterns
3. tdd-guide: Security test specifications

Pattern 4: Build Recovery

Agents: build-error-resolver → refactor-cleaner → tdd-guide

1. build-error-resolver: Fix immediate errors
2. refactor-cleaner: Clean up technical debt
3. tdd-guide: Add regression tests

---

Available Agents

Agent	Expertise	Trigger Phrases
architect	System Design	"design", "structure", "pattern"
planner	Task Planning	"plan", "roadmap", "breakdown"
code-reviewer	Code Quality	"review", "check", "audit"
security-reviewer	Security	"security", "auth", "vulnerability"
tdd-guide	Unit Testing	"test", "TDD", "coverage"
e2e-runner	E2E Testing	"e2e", "playwright", "integration"
build-error-resolver	Build Issues	"error", "compile", "typescript"
refactor-cleaner	Refactoring	"refactor", "clean", "improve"
doc-updater	Documentation	"docs", "readme", "document"
knowledge-agent	Knowledge	"learn", "pattern", "remember"

---

Synthesis Protocol

After all agents complete, synthesize findings:

## Orchestration Synthesis

### Task Summary

[What was accomplished]

### Agent Contributions

| Agent             | Finding         |
| ----------------- | --------------- |
| architect         | Designed X      |
| security-reviewer | Identified Y    |
| tdd-guide         | Created Z tests |

### Consolidated Deliverables

1. **Code Changes**: [summary]
2. **Test Coverage**: [metrics]
3. **Documentation**: [updates]

### Action Items

- [ ] Complete remaining tasks
- [ ] Run verification loop
- [ ] Update session state

---

Trust-Grade Integration

Pre-Orchestration Checklist

• Verify session state is current
• Check for active blockers
• Load relevant governance rules
• Confirm scope with user if complex

Post-Orchestration Protocol

• Run verification-loop skill
• Update session-state.json
• Trigger continuous-learning if patterns detected
• Document in knowledge base if novel approach

---

Best Practices

1. Logical Order: Discovery → Design → Implementation → Testing
2. Share Context: Pass relevant findings to subsequent agents
3. Single Synthesis: One unified report, not separate outputs
4. Verify Changes: Always include tdd-guide for code modifications
5. Maintain Governance: Apply Trust-Grade protocols throughout

---

Key Benefits

• ✅ Single session - All agents share context
• ✅ AI-controlled - Orchestrated autonomously
• ✅ Trust-Grade - Governance maintained throughout
• ✅ Resume support - Can continue previous agent work
• ✅ Context passing - Findings flow between agents