agent-coordination
Patterns for invoking and coordinating Automatasaurus agents. Use when delegating tasks to specialist agents.
When & Why to Use This Skill
This Claude skill provides a comprehensive framework for orchestrating specialized AI agents within the Automatasaurus ecosystem. It enables seamless task delegation, parallel execution, and quality gate management across developer, architect, designer, and tester roles, significantly optimizing the software development lifecycle through advanced multi-agent coordination.
Use Cases
- Automated PR Lifecycle Management: Coordinate a sequence of specialized agents to implement code, perform technical reviews, verify UI/UX compliance, and run automated tests before merging.
- Parallel Multi-Disciplinary Reviews: Invoke architect, designer, and tester agents simultaneously to review discovery plans or pull requests, reducing bottleneck times in the development pipeline.
- Intelligent Issue Escalation: Automatically route complex debugging tasks to an architect agent when a developer agent fails to resolve an issue after multiple attempts.
- Standardized Quality Gates: Implement rigorous approval workflows where specific agents must post standardized verification comments and pass quality checks before a task is considered complete.
| name | agent-coordination |
|---|---|
| description | Patterns for invoking and coordinating Automatasaurus agents. Use when delegating tasks to specialist agents. |
Agent Coordination Skill
This skill provides patterns for invoking the Automatasaurus agents.
Available Agents
| Agent | Purpose | When to Use |
|---|---|---|
developer |
Implements issues, creates PRs | When code needs to be written |
architect |
Technical reviews, analysis | For PR reviews, discovery reviews, stuck issues |
designer |
UI/UX reviews, design specs | For PR reviews, discovery reviews, design specs |
tester |
Testing, verification | For PR verification, running tests |
Invocation Patterns
Developer - Implement an Issue
Use the developer agent to implement issue #{number}.
Context:
- Issue: [title]
- Acceptance criteria: [from issue body]
- Design specs: [if applicable, from designer comments]
This is [MODE] mode. Create PR when implementation is complete.
Architect - Review Discovery Plan
Use the architect agent to review this discovery plan for technical feasibility.
Focus on:
- Architecture fit
- Scalability
- Security implications
- Technology choices
The discovery plan is at: [path to discovery.md]
Architect - Review PR
Use the architect agent to review PR #{pr_number} for technical quality.
Post a standardized approval comment when done:
- ✅ APPROVED - Architect
- OR ❌ CHANGES REQUESTED - Architect
Architect - Analyze Stuck Issue
Use the architect agent to analyze issue #{number}.
Developer has tried:
1. [attempt 1]
2. [attempt 2]
...
Error: [description]
Designer - Review Discovery Plan
Use the designer agent to review this discovery plan for UI/UX considerations.
Focus on:
- User flows
- Accessibility
- Responsive design
- Missing UI requirements
The discovery plan is at: [path to discovery.md]
Designer - Add Specs to Issue
Use the designer agent to add UI/UX specifications to issue #{number}.
Review the issue requirements and add design specs as a comment.
Designer - Review PR
Use the designer agent to review PR #{pr_number} for UI/UX quality.
If no UI changes, post: "**[Designer]** N/A - No UI changes in this PR."
Otherwise post standardized approval:
- ✅ APPROVED - Designer
- OR ❌ CHANGES REQUESTED - Designer
Tester - Verify PR
Use the tester agent to verify PR #{pr_number}.
Run tests and perform manual verification if needed.
Post a standardized approval comment when done:
- ✅ APPROVED - Tester
- OR ❌ CHANGES REQUESTED - Tester
Parallel Invocation
When reviews are independent, invoke agents in parallel:
# Invoke these in parallel (single message, multiple tool calls)
Use the architect agent to review PR #123
Use the designer agent to review PR #123
Use the tester agent to verify PR #123
Handoff Information
When delegating, include:
- Issue/PR number - What to work on
- Context - Acceptance criteria, design specs
- Mode - single-issue or all-issues (affects merge behavior)
- Expected output - What to produce (PR, comment, etc.)
Quality Gates
Each agent has quality gates:
| Agent | Gate |
|---|---|
| Developer | Tests passing, PR created |
| Architect | Technical review complete |
| Designer | Design review complete |
| Tester | All tests pass, verification complete |
Escalation Flow
Developer stuck (5 attempts)
↓
Architect analyzes
↓
If still stuck → Human escalation
.claude/hooks/request-attention.sh stuck "..."