Prompt Optimizer
Rewrites a vague prompt into a clear, testable one — adding role, context, constraints, output format, and success criteria — and explains each change.
What this skill does
Key features
- Adds an explicit role, context, and constraints
- Specifies the desired output format and success criteria
- Explains why each change was made
- Avoids over-engineering simple prompts
Use cases
- Sharpening a one-line prompt before running it
- Standardising prompts across a team
- Debugging a prompt that returns inconsistent output
SKILL.md
The skill definition — its metadata and the exact instructions an agent follows. Copy it into a SKILL.md file in your agent's skills folder.
| name | prompt-optimizer |
|---|---|
| description | Rewrite a vague prompt into a clear, testable one with role, context, constraints, format, and success criteria. |
| allowed-tools | Read, Write |
| output | rewritten prompt + change notes |
Musts
- Preserve the original intent; make it explicit rather than changing what is asked.
- Add a clear role, the relevant context, and any hard constraints.
- Specify the expected output format and measurable success criteria.
- Explain each change so the author can judge and adjust it.
- Flag ambiguity in the original instead of silently resolving it.
Guidelines
- Match effort to the task — do not bloat a simple prompt with ceremony.
- Prefer concrete, testable wording over vague adjectives like “good” or “detailed”.
- Keep constraints realistic; do not invent requirements the author never stated.
Output format
## Optimized prompt
<the rewritten, self-contained prompt>
## What changed and why
- Role: <added/adjusted> — <reason>
- Context: <added> — <reason>
- Constraints: <added> — <reason>
- Output format: <specified> — <reason>
- Success criteria: <specified> — <reason>
## Open questions
- <any ambiguity the author should resolve>FAQ
Does it make every prompt long and complex?
No. It adds only what improves clarity or testability. A simple prompt stays simple — over-engineering is called out as a failure mode, not a goal. The aim is the clearest prompt that gets a reliable result.
Will it change what I'm actually asking for?
No. It preserves your intent and only makes it explicit and testable. If the original intent is ambiguous, it flags the ambiguity and offers options rather than silently picking one.
Before running any skill, read its instructions and the tools it's allowed to use, and test it on a safe target first. See LLM security best practices.