1# Agents Index
2 
3Agents are reusable AI components with defined capabilities, tools, and instructions.
4 
5## Available Agents
6 
7### Evaluator Agent
8**Path**: `agents/evaluator-agent/evaluator-agent.md`
9**Purpose**: Assess the quality of LLM-generated responses
10 
11**Capabilities**:
12- Direct scoring against rubrics
13- Pairwise comparison of responses
14- Criteria extraction from task descriptions
15- Rubric generation for evaluation
16 
17**Tools Used**:
18- `directScore`
19- `pairwiseCompare`
20- `extractCriteria`
21- `generateRubric`
22 
23**Best For**:
24- Quality gates in content pipelines
25- Model comparison studies
26- RLHF preference data generation
27- Output validation before delivery
28 
29---
30 
31### Research Agent
32**Path**: `agents/research-agent/research-agent.md`
33**Purpose**: Gather, verify, and synthesize information from multiple sources
34 
35**Capabilities**:
36- Web search and result analysis
37- URL content extraction
38- Claim extraction and verification
39- Research synthesis
40 
41**Tools Used**:
42- `webSearch`
43- `readUrl`
44- `extractClaims`
45- `verifyClaim`
46- `synthesize`
47 
48**Best For**:
49- Knowledge base building
50- Fact checking
51- Market research
52- Technical documentation
53 
54---
55 
56### Orchestrator Agent
57**Path**: `agents/orchestrator-agent/orchestrator-agent.md`
58**Purpose**: Coordinate multi-agent workflows for complex tasks
59 
60**Capabilities**:
61- Task decomposition and assignment
62- Parallel task execution
63- Result synthesis
64- Error handling and recovery
65 
66**Tools Used**:
67- `delegateToAgent`
68- `parallelExecution`
69- `waitForCompletion`
70- `synthesizeResults`
71- `handleError`
72 
73**Best For**:
74- Complex multi-step tasks
75- Cross-capability workflows
76- Quality-assured pipelines
77- Long-running operations
78 
79## Agent Interaction Patterns
80 
81### Sequential Pipeline
82```
83Input → Agent A → Agent B → Agent C → Output
84```
85Use when each step depends on the previous.
86 
87### Parallel Fan-Out
88```
89        ┌→ Agent A ─┐
90Input ──┼→ Agent B ──┼→ Synthesis → Output
91        └→ Agent C ─┘
92```
93Use for independent subtasks that can run concurrently.
94 
95### Iterative Refinement
96```
97Input → Agent → Evaluator ─┬→ Output (if pass)
98                           └→ Agent (if fail, with feedback)
99```
100Use for quality-critical outputs.
101 
102## Adding New Agents
103 
1041. Create agent directory: `agents/<agent-name>/`
1052. Create main file: `agents/<agent-name>/<agent-name>.md`
1063. Define:
107   - Purpose and role
108   - System instructions
109   - Tool assignments
110   - Configuration options
111   - Usage examples
1124. Update this index
1135. Register with orchestrator if applicable
114 
115