| name | prompt-engineering |
| description | Use when designing or improving prompts, commands, hooks, skills, or sub-agent instructions, including output quality, reliability, and token efficiency. |
Prompt Engineering
Use this skill to design robust prompts that are clear, testable, and cost-aware.
When to Use This Skill
- Creating or refactoring prompts for skills, commands, hooks, and reusable templates
- Improving output consistency, formatting, or reasoning quality
- Reducing token usage or latency while preserving quality
- Choosing the right level of strictness (high, medium, low freedom)
- Defining guardrails and verification steps for fragile workflows
Workflow
- Define the exact task outcome, constraints, and required format.
- Choose the minimum instruction detail needed for the task's fragility.
- Start with a concise baseline prompt.
- Add structure: examples, explicit output schema, and failure handling.
- Add verification criteria for correctness and uncertainty handling.
- Test on varied and edge-case inputs, then iterate.
Expected Outputs
- A production-ready prompt (or template) with clear task, context, and output contract
- A short rationale for major design choices
- A test checklist with representative inputs and edge cases
- Notes on token and latency tradeoffs
References