1
---
2
name: agent-organization-expert
3
description: Multi-agent orchestration skill for team assembly, task decomposition, workflow optimization, and coordination strategies to achieve optimal team performance and resource utilization.
4
---
5
6
# Agent Organization
7
8
Assemble and coordinate multi-agent teams through systematic task analysis, capability mapping, and workflow design.
9
10
## Configuration
11
12
- **Agent Count**: ${agent_count:3}
13
- **Task Type**: ${task_type:general}
14
- **Orchestration Pattern**: ${orchestration_pattern:parallel}
15
- **Max Concurrency**: ${max_concurrency:5}
16
- **Timeout (seconds)**: ${timeout_seconds:300}
17
- **Retry Count**: ${retry_count:3}
18
19
## Core Process
20
21
1. **Analyze Requirements**: Understand task scope, constraints, and success criteria
22
2. **Map Capabilities**: Match available agents to required skills
23
3. **Design Workflow**: Create execution plan with dependencies and checkpoints
24
4. **Orchestrate Execution**: Coordinate ${agent_count:3} agents and monitor progress
25
5. **Optimize Continuously**: Adapt based on performance feedback
26
27
## Task Decomposition
28
29
### Requirement Analysis
30
- Break complex tasks into discrete subtasks
31
- Identify input/output requirements for each subtask
32
- Estimate complexity and resource needs per component
33
- Define clear success criteria for each unit
34
35
### Dependency Mapping
36
- Document task execution order constraints
37
- Identify data dependencies between subtasks
38
- Map resource sharing requirements
39
- Detect potential bottlenecks and conflicts
40
41
### Timeline Planning
42
- Sequence tasks respecting dependencies
43
- Identify parallelization opportunities (up to ${max_concurrency:5} concurrent)
44
- Allocate buffer time for high-risk components
45
- Define checkpoints for progress validation
46
47
## Agent Selection
48
49
### Capability Matching
50
Select agents based on:
51
- Required skills versus agent specializations
52
- Historical performance on similar tasks
53
- Current availability and workload capacity
54
- Cost efficiency for the task complexity
55
56
### Selection Criteria Priority
57
1. **Capability fit**: Agent must possess required skills
58
2. **Track record**: Prefer agents with proven success
59
3. **Availability**: Sufficient capacity for timely completion
60
4. **Cost**: Optimize resource utilization within constraints
61
62
### Backup Planning
63
- Identify alternate agents for critical roles
64
- Define failover triggers and handoff procedures
65
- Maintain redundancy for single-point-of-failure tasks
66
67
## Team Assembly
68
69
### Composition Principles
70
- Ensure complete skill coverage for all subtasks
71
- Balance workload across ${agent_count:3} team members
72
- Minimize communication overhead
73
- Include redundancy for critical functions
74
75
### Role Assignment
76
- Match agents to subtasks based on strength
77
- Define clear ownership and accountability
78
- Establish communication channels between dependent roles
79
- Document escalation paths for blockers
80
81
### Team Sizing
82
- Smaller teams for tightly coupled tasks
83
- Larger teams for parallelizable workloads
84
- Consider coordination overhead in sizing decisions
85
- Scale dynamically based on progress
86
87
## Orchestration Patterns
88
89
### Sequential Execution
90
Use when tasks have strict ordering requirements:
91
- Task B requires output from Task A
92
- State must be consistent between steps
93
- Error handling requires ordered rollback
94
95
### Parallel Processing
96
Use when tasks are independent (${orchestration_pattern:parallel}):
97
- No data dependencies between tasks
98
- Separate resource requirements
99
- Results can be aggregated after completion
100
- Maximum ${max_concurrency:5} concurrent operations
101
102
### Pipeline Pattern
103
Use for streaming or continuous processing:
104
- Each stage processes and forwards results
105
- Enables concurrent execution of different stages
106
- Reduces overall latency for multi-step workflows
107
108
### Hierarchical Delegation
109
Use for complex tasks requiring sub-orchestration:
110
- Lead agent coordinates sub-teams
111
- Each sub-team handles a domain
112
- Results aggregate upward through hierarchy
113
114
### Map-Reduce
115
Use for large-scale data processing:
116
- Map phase distributes work across agents
117
- Each agent processes a partition
118
- Reduce phase combines results
119
120
## Workflow Design
121
122
### Process Structure
123
1. **Entry point**: Validate inputs and initialize state
124
2. **Execution phases**: Ordered task groupings
125
3. **Checkpoints**: State persistence and validation points
126
4. **Exit point**: Result aggregation and cleanup
127
128
### Control Flow
129
- Define branching conditions for alternative paths
130
- Specify retry policies for transient failures (max ${retry_count:3} retries)
131
- Establish timeout thresholds per phase (${timeout_seconds:300}s default)
132
- Plan graceful degradation for partial failures
133
134
### Data Flow
135
- Document data transformations between stages
136
- Specify data formats and validation rules
137
- Plan for data persistence at checkpoints
138
- Handle data cleanup after completion
139
140
## Coordination Strategies
141
142
### Communication Patterns
143
- **Direct**: Agent-to-agent for tight coupling
144
- **Broadcast**: One-to-many for status updates
145
- **Queue-based**: Asynchronous for decoupled tasks
146
- **Event-driven**: Reactive to state changes
147
148
### Synchronization
149
- Define sync points for dependent tasks
150
- Implement waiting mechanisms with timeouts (${timeout_seconds:300}s)
151
- Handle out-of-order completion gracefully
152
- Maintain consistent state across agents
153
154
### Conflict Resolution
155
- Establish priority rules for resource contention
156
- Define arbitration mechanisms for conflicts
157
- Document rollback procedures for deadlocks
158
- Prevent conflicts through careful scheduling
159
160
## Performance Optimization
161
162
### Load Balancing
163
- Distribute work based on agent capacity
164
- Monitor utilization and rebalance dynamically
165
- Avoid overloading high-performing agents
166
- Consider agent locality for data-intensive tasks
167
168
### Bottleneck Management
169
- Identify slow stages through monitoring
170
- Add capacity to constrained resources
171
- Restructure workflows to reduce dependencies
172
- Cache intermediate results where beneficial
173
174
### Resource Efficiency
175
- Pool shared resources across agents
176
- Release resources promptly after use
177
- Batch similar operations to reduce overhead
178
- Monitor and alert on resource waste
179
180
## Monitoring and Adaptation
181
182
### Progress Tracking
183
- Monitor completion status per task
184
- Track time spent versus estimates
185
- Identify tasks at risk of delay
186
- Report aggregated progress to stakeholders
187
188
### Performance Metrics
189
- Task completion rate and latency
190
- Agent utilization and throughput
191
- Error rates and recovery times
192
- Resource consumption and cost
193
194
### Dynamic Adjustment
195
- Reallocate agents based on progress
196
- Adjust priorities based on blockers
197
- Scale team size based on workload
198
- Modify workflow based on learning
199
200
## Error Handling
201
202
### Failure Detection
203
- Monitor for task failures and timeouts (${timeout_seconds:300}s threshold)
204
- Detect agent unavailability promptly
205
- Identify cascade failure patterns
206
- Alert on anomalous behavior
207
208
### Recovery Procedures
209
- Retry transient failures with backoff (up to ${retry_count:3} attempts)
210
- Failover to backup agents when needed
211
- Rollback to last checkpoint on critical failure
212
- Escalate unrecoverable issues
213
214
### Prevention
215
- Validate inputs before execution
216
- Test agent availability before assignment
217
- Design for graceful degradation
218
- Build redundancy into critical paths
219
220
## Quality Assurance
221
222
### Validation Gates
223
- Verify outputs at each checkpoint
224
- Cross-check results from parallel tasks
225
- Validate final aggregated results
226
- Confirm success criteria are met
227
228
### Performance Standards
229
- Agent selection accuracy target: >${agent_selection_accuracy:95}%
230
- Task completion rate target: >${task_completion_rate:99}%
231
- Response time target: <${response_time_threshold:5} seconds
232
- Resource utilization: optimal range ${utilization_min:60}-${utilization_max:80}%
233
234
## Best Practices
235
236
### Planning
237
- Invest time in thorough task analysis
238
- Document assumptions and constraints
239
- Plan for failure scenarios upfront
240
- Define clear success metrics
241
242
### Execution
243
- Start with minimal viable team (${agent_count:3} agents)
244
- Scale based on observed needs
245
- Maintain clear communication channels
246
- Track progress against milestones
247
248
### Learning
249
- Capture performance data for analysis
250
- Identify patterns in successes and failures
251
- Refine selection and coordination strategies
252
- Share learnings across future orchestrations