prompt-optimizer
Optimize system prompts for Claude Code agents using proven prompt engineering patterns. Use when users request prompt improvement, optimization, or refinement for agent workflows, tool instructions, or system behaviors.
When & Why to Use This Skill
The Prompt Engineer skill is a specialized utility designed to optimize system prompts for Claude Code agents using industry-proven prompt engineering patterns. It automates the refinement of agent workflows, tool instructions, and system behaviors, enabling developers to enhance AI performance through structured optimization scopes including single-prompt tuning, ecosystem management, and greenfield prompt design.
Use Cases
- Single-Prompt Optimization: Refining a specific system prompt to improve instruction following, tool-calling precision, and response quality.
- Ecosystem Alignment: Harmonizing multiple related prompts within a complex agentic workflow to ensure consistent behavior and seamless interaction.
- Greenfield Prompt Design: Creating high-quality system prompts from scratch based on initial project requirements and functional goals.
- Troubleshooting & Debugging: Identifying and fixing specific behavioral issues or logic gaps in existing prompts to resolve performance bottlenecks.
| name | prompt-engineer |
|---|---|
| description | Invoke IMMEDIATELY via python script when user requests prompt optimization. Do NOT analyze first - invoke this skill immediately. |
Prompt Engineer
When this skill activates, IMMEDIATELY invoke the script. The script IS the workflow.
Invocation
Start with step 1 (triage) to determine scope:
Then continue with determined scope:
| Argument | Required | Description |
|---|---|---|
--step |
Yes | Current step (1 = triage, 2-6 = workflow) |
--scope |
For 2+ | Required for steps 2-6. Determined by step 1. |
Scopes
- single-prompt: One prompt file, general optimization
- ecosystem: Multiple related prompts that interact
- greenfield: No existing prompt, designing from requirements
- problem: Existing prompt(s) with specific issue to fix
Do NOT analyze or explore first. Run the script and follow its output.