@cloudkinetix/bmad-enhanced/expansion-packs/ck-llm-agent-dev/templates/llm-agent-spec-tmpl.md

Cloud-Kinetix enhanced fork of BMAD-METHOD - Breakthrough Method of Agile AI-driven Development with robust versioning and unified validation.

# {{Agent Name}} LLM Agent Specification

[[LLM: Dynamic Specification Generation

1. **Context Analysis**: Understand the specific project context, requirements, and constraints
2. **Research Current Standards**: Research latest best practices for LLM agent specifications
3. **Adaptive Template**: Customize this template based on analyzed needs and current industry standards
4. **Technology-Aware**: Consider current LLM capabilities and limitations when setting requirements
5. **Requirements Elicitation**: Use systematic elicitation to uncover both explicit and implicit needs
6. **Future-Proofing**: Design specifications that can evolve with advancing LLM capabilities

**Research Before Starting**:

- What are current best practices for LLM agent specifications in this domain?
- What recent developments in LLM should influence the design requirements?
- What are the latest safety and compliance considerations?
- How do successful similar agents approach the specified use case?

Output file location: `docs/llm-agents/{{agent-name}}-spec.md`]]

## Agent Overview

[[LLM: Provide a comprehensive overview of the LLM agent, including its purpose, target users, and value proposition. Be specific about the problems it solves.]]

- **Agent Name**: {{agent-name}}
- **Agent Type**: {{Conversational/Task-oriented/Autonomous/Multi-modal}}
- **Primary Purpose**: {{main-purpose}}
- **Target Users**: {{user-segments}}
- **Key Capabilities**: {{list 3-5 core capabilities}}

## Functional Requirements

[[LLM: Detail what the agent must do. Be specific about inputs, outputs, and behaviors.]]

### Core Capabilities

<<REPEAT: capability>>

#### {{Capability Name}}

- **Description**: {{detailed-description}}
- **Input Types**: {{expected-inputs}}
- **Output Format**: {{output-structure}}
- **Success Criteria**: {{measurable-criteria}}
- **Error Handling**: {{error-scenarios}}
  <</REPEAT>>

### Conversation Flow

[[LLM: For conversational agents, define the interaction patterns and dialogue management.]]

- **Greeting/Initialization**: {{initial-interaction}}
- **Context Management**: {{how-context-is-maintained}}
- **Turn-Taking**: {{conversation-rules}}
- **Closure/Handoff**: {{end-of-interaction}}

## Non-Functional Requirements

### Performance Requirements

[[LLM: Define specific, measurable performance targets.]]

- **Response Time**: {{latency-requirements}}
- **Throughput**: {{requests-per-second}}
- **Availability**: {{uptime-percentage}}
- **Scalability**: {{scaling-requirements}}
- **Resource Limits**: {{cpu/memory/cost-constraints}}

### Safety and Alignment

[[LLM: Detail safety measures and alignment strategies.]]

- **Content Filtering**: {{harmful-content-prevention}}
- **Bias Mitigation**: {{fairness-measures}}
- **Privacy Protection**: {{data-handling-policies}}
- **Security Measures**: {{injection-prevention}}
- **Alignment Constraints**: {{behavioral-boundaries}}

## Technical Architecture

### Model Selection

[[LLM: Research and document model choices with current analysis and rationale.]]

**Research Process**:

- Research current available models and their capabilities for this use case
- Analyze performance benchmarks relevant to the agent's requirements
- Evaluate cost implications across different model options
- Consider integration complexity and API stability

**Model Strategy**:

- **Primary Model**: {{model-name-version}} - Selected based on {{research-based-rationale}}
- **Fallback Models**: {{backup-options}} - Chosen for {{fallback-reasoning}}
- **Selection Criteria**: {{research-informed-criteria}}
- **Fine-tuning Requirements**: {{customization-needs-analysis}}
- **Cost Considerations**: {{budget-constraints-and-optimization}}

### Integration Points

[[LLM: Define how the agent connects with other systems.]]

- **APIs/Services**: {{external-dependencies}}
- **Data Sources**: {{knowledge-bases}}
- **Authentication**: {{auth-methods}}
- **Rate Limiting**: {{throttling-policies}}
- **Error Recovery**: {{retry-strategies}}

## Prompt Engineering Strategy

### System Prompts

[[LLM: Define the core system prompts that shape agent behavior.]]

```text
{{system-prompt-template}}
```

### Prompt Patterns

- **Task Decomposition**: {{how-complex-tasks-are-handled}}
- **Context Injection**: {{dynamic-context-management}}
- **Output Formatting**: {{structured-response-patterns}}
- **Error Handling**: {{graceful-failure-prompts}}

### Testing Strategy

[[LLM: Research current testing methodologies and design appropriate testing strategy.]]

**Research Areas**:

- Current best practices for LLM agent testing and evaluation
- Available testing frameworks and their suitability for this use case
- Industry-standard evaluation metrics for similar agents
- Latest developments in prompt testing and optimization

**Testing Approach**:

- **Test Coverage**: {{research-informed-test-scenarios}}
- **Evaluation Metrics**: {{current-best-practice-measures}}
- **Testing Framework**: {{selected-framework-with-rationale}}
- **Optimization Strategy**: {{evidence-based-optimization-approach}}

## Monitoring and Observability

### Key Metrics

[[LLM: Define what will be measured and monitored.]]

- **Quality Metrics**: {{accuracy/relevance/completeness}}
- **Performance Metrics**: {{latency/throughput/errors}}
- **Business Metrics**: {{user-satisfaction/engagement}}
- **Cost Metrics**: {{per-request/daily-budget}}

### Alerting Thresholds

<<REPEAT: alert>>

- **{{Metric Name}}**:
  - Warning: {{warning-threshold}}
  - Critical: {{critical-threshold}}
  - Action: {{remediation-steps}}
    <</REPEAT>>

## Development Phases

### Phase 1: MVP

- **Timeline**: {{duration}}
- **Features**: {{core-features-only}}
- **Success Criteria**: {{mvp-goals}}

### Phase 2: Enhancement

- **Timeline**: {{duration}}
- **Features**: {{additional-capabilities}}
- **Success Criteria**: {{enhancement-goals}}

### Phase 3: Scale

- **Timeline**: {{duration}}
- **Features**: {{scale-features}}
- **Success Criteria**: {{scale-goals}}

## Risk Assessment

[[LLM: Identify and plan for potential risks.]]

<<REPEAT: risk>>

### {{Risk Name}}

- **Probability**: {{High/Medium/Low}}
- **Impact**: {{High/Medium/Low}}
- **Mitigation**: {{prevention-strategy}}
- **Contingency**: {{if-it-happens}}
  <</REPEAT>>

## Testing and Validation

### Test Scenarios

[[LLM: Define comprehensive test cases covering normal and edge cases.]]

1. **Functional Tests**: {{core-functionality-tests}}
2. **Edge Cases**: {{boundary-conditions}}
3. **Adversarial Tests**: {{security-tests}}
4. **Performance Tests**: {{load-testing}}
5. **Integration Tests**: {{system-integration}}

### Acceptance Criteria

- [ ] All functional requirements met
- [ ] Performance targets achieved
- [ ] Safety measures validated
- [ ] Integration tests passing
- [ ] User acceptance confirmed

## Documentation Requirements

- **API Documentation**: {{api-spec-location}}
- **User Guide**: {{end-user-docs}}
- **Operations Manual**: {{ops-runbook}}
- **Integration Guide**: {{developer-docs}}

## Compliance and Legal

- **Data Protection**: {{gdpr/ccpa-compliance}}
- **Industry Standards**: {{relevant-standards}}
- **Audit Requirements**: {{logging-needs}}
- **Terms of Service**: {{usage-policies}}

## Success Metrics

[[LLM: Define how success will be measured post-deployment.]]

- **Adoption Metrics**: {{user-growth-targets}}
- **Quality Metrics**: {{satisfaction-scores}}
- **Business Metrics**: {{roi-targets}}
- **Technical Metrics**: {{reliability-goals}}

## Next Steps

1. Review and approve specification
2. Create detailed technical design
3. Set up development environment
4. Implement MVP features
5. Configure testing framework
6. Plan deployment strategy