prompt-optimizer-skill

内容来源：README.md（说明文档） · 原始地址 · 查看安装指南

原始内容

# Claude Prompt Optimizer Skill

> Interpreter that enhances underspecified prompts before execution, producing better outputs directly.

[![License: MIT](https://img.shields.io/badge/License-MIT-111827?style=for-the-badge&logo=github)](https://opensource.org/licenses/MIT)
[![Version](https://img.shields.io/badge/Version-3.0-2563EB?style=for-the-badge)](https://github.com/joaocarlos/prompt-optimizer-skill/releases)
[![Claude Skill](https://img.shields.io/badge/Claude_Skill-7C3AED?style=for-the-badge&logo=anthropic)](https://claude.ai)

## What It Does

Prompt Optimizer works as a silent preprocessing layer that analyzes your input, infers missing context and specifications, then executes an enhanced version to produce better results directly.

**Example:**

- **You say**: "Draft an article outline based on project specifications"
- **Claude delivers**: A detailed hierarchical outline with section descriptions, key points, logical flow, and estimated proportions - not a lecture about prompt quality

The skill optimizes *by interpreting and executing*, not by outputting optimized prompt text.

## Key Features

- **Silent preprocessing** - Enhances your input without interrupting workflow
- **Direct results** - Outputs the actual result, not meta-commentary about prompts
- **Context-aware** - Uses project files, conversation history, and domain knowledge
- **Transparent when needed** - Briefly states interpretation for course-correction
- **Universal coverage** - Works across writing, code, research, planning, and more

## Installation

### Claude Desktop

1. [Download prompt-optimizer.skill](https://github.com/joaocarlos/prompt-optimizer-skill/releases/latest/download/prompt-optimizer.skill) from the latest release
2. Open Claude Desktop
3. Go to **Settings > Skills**
4. Click **Upload skill**
5. Select the downloaded `prompt-optimizer.skill` file
6. Confirm installation

### Claude Code

**User-level** (applies to all projects):

```bash
cp -r skill ~/.claude/skills/prompt-optimizer
```

**Project-level** (applies to current project only):

```bash
cp -r skill .claude/skills/prompt-optimizer
```

## How It Works

### The 4-Step Process

1. **Analyze** - Scan input for missing specifications (format, depth, audience, purpose, tone, constraints)
2. **Infer** - Make intelligent assumptions based on context clues and domain conventions
3. **Execute** - Process as if you had provided all specifications
4. **Clarify** - Briefly state interpretation when it would help (optional)

### When It Applies

**Applies to tasks that:**

- Lack specificity about format, structure, or depth
- Miss context about audience or purpose
- Use general terms that could mean many things
- Would benefit from reasonable assumptions

**Does NOT apply to:**

- Simple factual questions
- Already well-specified prompts
- Exploratory/brainstorming tasks
- Cases where minimal interpretation is preferred

## Usage Examples

### Writing Task

**You say**: "Draft an article outline based on project specifications"

**Claude responds**:

```text
Interpreting this as a detailed hierarchical outline with section descriptions,
key points, and logical flow.

## Article Outline: [Title]

### 1. Introduction (10%)
- Hook: [specific angle]
- Context: [background]
- Thesis: [main argument]
...
```

### Code Task

**You say**: "Create a function to process the data"

**Claude responds**:

```text
Interpreting this as a Python function matching your project's style, with
type hints, error handling, and documentation.
```

```python
def process_data(records: list[dict]) -> ProcessedResult:
    """Process raw records into analyzed format."""
    ...
```

### When NOT Applied

**You say**: "What's the difference between TCP and UDP?"

**Claude responds** (directly, no interpretation prefix):

```text
TCP is connection-oriented and guarantees delivery through acknowledgments.
UDP is connectionless and faster but doesn't guarantee delivery...
```

## Domain-Specific Defaults

| Domain | Default Inferences |
|--------|-------------------|
| **Writing** | Length matches complexity, clear sections, professional tone, knowledgeable audience |
| **Code** | Infer language from context, follow codebase conventions, include error handling and types |
| **Research** | Comprehensive depth, cite evidence, executive summary + findings + conclusions |
| **Planning** | Cover immediate + long-term, structured phases, actionable steps, acknowledge tradeoffs |

## Calibration

### Infer More When

- Input is very short for a complex task
- Task type has clear professional standards
- Previous context suggests comprehensive outputs wanted

### Infer Less When

- User has been very specific previously
- Task is exploratory or creative
- User says "quick" or "brief"

### Ask Instead When

- Critical ambiguity that could waste significant effort
- Multiple valid interpretations with very different outputs
- Missing information that can't be reasonably inferred

## Version History

### v3.0 (Current)

- Complete redesign as silent preprocessing interpreter
- Produces results directly instead of outputting optimized prompts
- Context-aware inference using project files and conversation history
- Optional interpretation statements for transparency
- Calibration guidelines for when to infer more/less

### v2.0

- Proactive prompt detection
- Universal domain coverage
- Automatic triggering

### v1.0

- Initial release

## Compatibility

- **Platform**: Claude Code, Claude.ai (with skills support)
- **Format**: Standard Claude Skill (SKILL.md)
- **Dependencies**: None (self-contained)

## Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

### Reporting Issues

Found a bug or have a suggestion? [Open an issue](https://github.com/joaocarlos/prompt-optimizer-skill/issues/new)

## License

MIT License - see [LICENSE](https://github.com/joaocarlos/prompt-optimizer-skill/tree/HEAD/LICENSE) for details.

## Author

Created by Joao Carlos N. Bittencourt - [GitHub](https://github.com/joaocarlos)

---

**Ready to start?** Copy the skill to your Claude skills directory and get better outputs automatically.

Claude Prompt Optimizer Skill

Interpreter that enhances underspecified prompts before execution, producing better outputs directly.

What It Does

Prompt Optimizer works as a silent preprocessing layer that analyzes your input, infers missing context and specifications, then executes an enhanced version to produce better results directly.

Example:

You say: "Draft an article outline based on project specifications"
Claude delivers: A detailed hierarchical outline with section descriptions, key points, logical flow, and estimated proportions - not a lecture about prompt quality

The skill optimizes by interpreting and executing, not by outputting optimized prompt text.

Key Features

Silent preprocessing - Enhances your input without interrupting workflow
Direct results - Outputs the actual result, not meta-commentary about prompts
Context-aware - Uses project files, conversation history, and domain knowledge
Transparent when needed - Briefly states interpretation for course-correction
Universal coverage - Works across writing, code, research, planning, and more

Installation

Claude Desktop

Download prompt-optimizer.skill from the latest release
Open Claude Desktop
Go to Settings > Skills
Click Upload skill
Select the downloaded prompt-optimizer.skill file
Confirm installation

Claude Code

User-level (applies to all projects):

cp -r skill ~/.claude/skills/prompt-optimizer

Project-level (applies to current project only):

cp -r skill .claude/skills/prompt-optimizer

How It Works

The 4-Step Process

Analyze - Scan input for missing specifications (format, depth, audience, purpose, tone, constraints)
Infer - Make intelligent assumptions based on context clues and domain conventions
Execute - Process as if you had provided all specifications
Clarify - Briefly state interpretation when it would help (optional)

When It Applies

Applies to tasks that:

Lack specificity about format, structure, or depth
Miss context about audience or purpose
Use general terms that could mean many things
Would benefit from reasonable assumptions

Does NOT apply to:

Simple factual questions
Already well-specified prompts
Exploratory/brainstorming tasks
Cases where minimal interpretation is preferred

Usage Examples

Writing Task

You say: "Draft an article outline based on project specifications"

Claude responds:

Interpreting this as a detailed hierarchical outline with section descriptions,
key points, and logical flow.

## Article Outline: [Title]

### 1. Introduction (10%)
- Hook: [specific angle]
- Context: [background]
- Thesis: [main argument]
...

Code Task

You say: "Create a function to process the data"

Claude responds:

Interpreting this as a Python function matching your project's style, with
type hints, error handling, and documentation.

def process_data(records: list[dict]) -> ProcessedResult:
    """Process raw records into analyzed format."""
    ...

When NOT Applied

You say: "What's the difference between TCP and UDP?"

Claude responds (directly, no interpretation prefix):

TCP is connection-oriented and guarantees delivery through acknowledgments.
UDP is connectionless and faster but doesn't guarantee delivery...

Domain-Specific Defaults

Domain	Default Inferences
Writing	Length matches complexity, clear sections, professional tone, knowledgeable audience
Code	Infer language from context, follow codebase conventions, include error handling and types
Research	Comprehensive depth, cite evidence, executive summary + findings + conclusions
Planning	Cover immediate + long-term, structured phases, actionable steps, acknowledge tradeoffs

Calibration

Infer More When

Input is very short for a complex task
Task type has clear professional standards
Previous context suggests comprehensive outputs wanted

Infer Less When

User has been very specific previously
Task is exploratory or creative
User says "quick" or "brief"

Ask Instead When

Critical ambiguity that could waste significant effort
Multiple valid interpretations with very different outputs
Missing information that can't be reasonably inferred

Version History

v3.0 (Current)

Complete redesign as silent preprocessing interpreter
Produces results directly instead of outputting optimized prompts
Context-aware inference using project files and conversation history
Optional interpretation statements for transparency
Calibration guidelines for when to infer more/less

v2.0

Proactive prompt detection
Universal domain coverage
Automatic triggering

v1.0

Initial release

Compatibility

Platform: Claude Code, Claude.ai (with skills support)
Format: Standard Claude Skill (SKILL.md)
Dependencies: None (self-contained)

Contributing

Contributions welcome! Please:

Fork the repository
Create a feature branch
Make your changes
Submit a pull request

Reporting Issues

Found a bug or have a suggestion? Open an issue

License

MIT License - see LICENSE for details.

Author

Created by Joao Carlos N. Bittencourt - GitHub

Ready to start? Copy the skill to your Claude skills directory and get better outputs automatically.