sharaf/agent-prompt-engineer
Write or audit AI agent system prompts component-by-component across identity, instruction architecture, behavioral constraints, tools, examples, context strategy, output format, and error handling. Use when the user wants to design a new agent prompt, write a system prompt, review an existing agent prompt, fix tool-use instructions, audit prompt structure, improve context strategy, tune output formats, or define error handling for single-agent or multi-agent systems.
100
100%
Does it follow best practices?
Impact
100%
1.33xAverage score across 3 eval scenarios
Passed
No known issues
Evaluation results
100%
32%Agent Prompt Design: Customer Support Triage Agent
Write-mode system prompt for a customer support agent
First Actions section
0%
100%
Design Decisions section
87%
100%
System Prompt section
100%
100%
Scope boundaries
100%
100%
Tool definitions
70%
100%
Escalation triggers
100%
100%
Tool failure handling
100%
100%
Context strategy
37%
100%
Output format
100%
100%
Critical rule sandwich
37%
100%
How to Iterate
12%
100%
Positive framing
66%
100%
100%
35%Agent Prompt Audit: Research Assistant
Audit-mode review of an existing agent system prompt
Audit title and First Actions
40%
100%
Component Scores
10%
100%
Highest-impact fixes
100%
100%
Specific evidence
80%
100%
Concrete rewrite blocks
100%
100%
Tool definition diagnosis
25%
100%
Retrieval failure handling
100%
100%
Output format diagnosis
37%
100%
Constraint framing
87%
100%
Priority labels
100%
100%
Open Questions
0%
100%
No vague findings
100%
100%
100%
7%Multi-Agent Prompt Design: Sales Intelligence Workflow
Multi-agent prompt architecture for a sales intelligence workflow
First Actions and mode
37%
100%
Agent-specific prompts
100%
100%
Role boundaries
100%
100%
Orchestrator constraints
100%
100%
Worker contracts
100%
100%
Verifier rubric
100%
100%
Typed handoff schema
100%
100%
Context isolation
87%
100%
Unsupported claim handling
100%
100%
Error handling
100%
100%
Implementation-ready output
100%
100%
How to Iterate
75%
100%
- Evaluated
- Agent
- Claude
- Model
- Claude Sonnet 4.6