1# Prompts Index
2 
3Prompts are reusable templates that define how agents and tools interact with LLMs.
4 
5## Prompt Categories
6 
7### Evaluation Prompts
8**Path**: `prompts/evaluation/`
9 
10Templates for quality assessment tasks.
11 
12| Prompt | Purpose | Used By |
13|--------|---------|---------|
14| `direct-scoring-prompt` | Evaluate single response | Evaluator Agent, directScore tool |
15| `pairwise-comparison-prompt` | Compare two responses | Evaluator Agent, pairwiseCompare tool |
16 
17---
18 
19### Research Prompts
20**Path**: `prompts/research/`
21 
22Templates for information gathering and synthesis.
23 
24| Prompt | Purpose | Used By |
25|--------|---------|---------|
26| `research-synthesis-prompt` | Synthesize findings | Research Agent |
27 
28---
29 
30### Agent System Prompts
31**Path**: `prompts/agent-system/`
32 
33System prompts for agent definitions.
34 
35| Prompt | Purpose | Used By |
36|--------|---------|---------|
37| `orchestrator-prompt` | Multi-agent coordination | Orchestrator Agent |
38 
39## Prompt Template Format
40 
41### Standard Structure
42 
43```markdown
44# Prompt Name
45 
46## Purpose
47Brief description of what this prompt accomplishes.
48 
49## Prompt Template
50```markdown
51[The actual prompt with {{variables}}]
52```
53 
54## Variables
55| Variable | Description | Required |
56|----------|-------------|----------|
57| var_name | What it contains | Yes/No |
58 
59## Example Usage
60Concrete example showing inputs and expected outputs.
61 
62## Best Practices
63Guidelines for using this prompt effectively.
64```
65 
66### Variable Syntax
67 
68Use Handlebars-style templating:
69 
70```markdown
71{{variable}}                 # Simple substitution
72{{#if condition}}...{{/if}} # Conditional section
73{{#each array}}...{{/each}} # Iteration
74```
75 
76## Prompt Design Principles
77 
78### 1. Clear Role Definition
79Tell the model exactly what it is and what it's doing.
80 
81```markdown
82You are an expert evaluator assessing the quality of AI-generated responses.
83```
84 
85### 2. Explicit Instructions
86Don't assume the model will infer requirements.
87 
88```markdown
89For each criterion:
901. First, identify specific evidence from the response
912. Then, determine the appropriate score based on the rubric
923. Finally, provide actionable feedback
93```
94 
95### 3. Structured Output
96Specify the exact format you need.
97 
98```markdown
99Format your response as structured JSON:
100```json
101{
102  "scores": [...],
103  "summary": {...}
104}
105```
106```
107 
108### 4. Guard Rails
109Include constraints and warnings.
110 
111```markdown
112Important Guidelines:
113- Do NOT prefer responses simply because they are longer
114- Do NOT prefer responses based on their position (A vs B)
115- Focus on the specified criteria
116```
117 
118## Adding New Prompts
119 
1201. Determine category or create new: `prompts/<category>/`
1212. Create prompt file: `prompts/<category>/<prompt-name>.md`
1223. Include:
123   - Purpose
124   - Template with variables
125   - Variable documentation
126   - Example usage
127   - Best practices
1284. Update this index
129 
130## Prompt Testing Checklist
131 
132- [ ] Variables render correctly
133- [ ] Output format is parseable
134- [ ] Edge cases are handled
135- [ ] Instructions are unambiguous
136- [ ] Examples match expected output
137- [ ] Constraints are clear
138 
139