0
Improve prompt
Bir yapay zeka istemini netlik, spesifiklik ve etkililik açısından analiz edin ve iyileştirin.
yazan kuscubasiece·8 Eylül 2026
Kaynak: rohitg00/awesome-claude-code-toolkit · Apache-2.0
Bir yapay zeka istemini netlik, spesifiklik ve etkililik açısından analiz edin ve iyileştirin.
Kaynak: rohitg00/awesome-claude-code-toolkit · Apache-2.0
mkdir -p .claude/commands && curl -sL https://www.antrophiclove.com/api/kaynak/improve-prompt/raw -o .claude/commands/improve-prompt.mdHedef dosya: .claude/commands/improve-prompt.md
---
name: improve-prompt
description: Analyze and improve an AI prompt for clarity, specificity, and effectiveness.
---
Analyze and improve an AI prompt for clarity, specificity, and effectiveness.
## Steps
1. Read the provided prompt or command definition.
2. Evaluate against prompt quality dimensions:
- **Clarity**: Is the instruction unambiguous?
- **Specificity**: Does it define the expected output format?
- **Context**: Does it provide enough background information?
- **Constraints**: Are boundaries and rules clearly stated?
- **Examples**: Are there concrete examples of good output?
3. Identify common issues:
- Vague verbs ("handle", "process", "manage") without specifics.
- Missing edge case guidance.
- No output format specification.
- Contradictory instructions.
- Missing success criteria.
4. Rewrite the prompt with improvements:
- Add structured sections (Steps, Format, Rules).
- Replace vague instructions with specific actions.
- Add output format with examples.
- Define explicit constraints and error handling.
5. Compare original vs improved versions side by side.
## Format
```
Prompt Analysis:
Clarity: <1-5>/5
Specificity: <1-5>/5
Context: <1-5>/5
Issues found:
- <issue description>
Improved prompt:
<rewritten prompt>
Changes made:
- <what was improved and why>
```
## Rules
- Preserve the original intent; only improve clarity and structure.
- Use imperative mood for instructions ("Run X" not "You should run X").
- Every prompt should have Steps, Format, and Rules sections.
- Limit prompts to under 50 lines; split complex workflows into sub-commands.
- Test the improved prompt against the original to verify better results.