Find your landing page message
Connect a product’s value to a specific audience and write a clear first-screen message.
Build custom AI prompts from scratch using guided workflows and patterns from a 2,500+ prompt database. Supports image, video, and text prompts for Midjourney, Flux, Leonardo AI, DALL-E, Sora, and 12 more models. Use when user says "build a prompt", "create a prompt", "write a prompt", "make me a...
---
name: prompt-build-5026f45145f2
description: >
Build custom AI prompts from scratch using guided workflows and patterns
from a 2,500+ prompt database. Supports image, video, and text prompts
for Midjourney, Flux, Leonardo AI, DALL-E, Sora, and 12 more models.
Use when user says "build a prompt", "create a prompt", "write a prompt",
"make me a prompt", "new prompt", or "prompt from scratch".
---
# Prompt Builder
Build production-quality AI prompts using patterns extracted from 2,500+ curated prompts.
## Build Workflow
### Step 1: Determine Intent
Ask the user (if not already clear):
1. **Output type**: Image, Video, or Text?
2. **Target model**: Which AI model? (Midjourney, Flux, DALL-E, Leonardo, Sora, etc.)
3. **Subject**: What should the prompt create?
4. **Style/mood**: Any specific aesthetic?
### Step 2: Find Reference Prompts
Search the database for similar prompts to use as inspiration:
```bash
python3 {PROMPT_ENGINE_DIR}/scripts/search_prompts.py "SUBJECT" --model MODEL --limit 5
```
### Step 3: Construct the Prompt
Use this universal structure (adapt per model):
**Image Prompts:**
```
[Subject/Action] + [Setting/Environment] + [Lighting/Mood] + [Camera/Lens] +
[Style Reference] + [Technical Parameters]
```
**Video Prompts:**
```
[Camera Movement] + [Subject/Action] + [Setting] + [Lighting] +
[Duration/Speed] + [Style] + [Technical]
```
**Text Prompts:**
```
[Role/Context] + [Task] + [Constraints] + [Format] + [Tone] + [Examples]
```
### Step 4: Apply Model-Specific Syntax
Load `{PROMPT_ENGINE_DIR}/references/model-guide.md` for model-specific formatting rules:
- **Midjourney**: Use `--ar`, `--v`, `--style`, `--s`, `--chaos`
- **Flux**: Natural language, detailed descriptions work best
- **DALL-E**: Direct, descriptive, avoid negative prompts
- **Leonardo AI**: Use model-specific keywords (Phoenix, Alchemy)
- **Sora**: Camera movements, scene transitions, temporal descriptions
### Step 5: Review and Refine
Present the prompt with:
1. The full prompt text
2. Recommended parameters/settings
3. Similar prompts from the database for comparison
4. Suggested variations (3 alternatives)
## Quality Gates
- Prompt must be > 30 characters
- Must include subject + at least one modifier (style, lighting, or camera)
- Must use model-appropriate syntax
- No conflicting instructions (e.g., "realistic" + "cartoon")
---
name: prompt-build-5026f45145f2
description: >
Build custom AI prompts from scratch using guided workflows and patterns
from a 2,500+ prompt database. Supports image, video, and text prompts
for Midjourney, Flux, Leonardo AI, DALL-E, Sora, and 12 more models.
Use when user says "build a prompt", "create a prompt", "write a prompt",
"make me a prompt", "new prompt", or "prompt from scratch".
---
# Prompt Builder
Build production-quality AI prompts using patterns extracted from 2,500+ curated prompts.
## Build Workflow
### Step 1: Determine Intent
Ask the user (if not already clear):
1. **Output type**: Image, Video, or Text?
2. **Target model**: Which AI model? (Midjourney, Flux, DALL-E, Leonardo, Sora, etc.)
3. **Subject**: What should the prompt create?
4. **Style/mood**: Any specific aesthetic?
### Step 2: Find Reference Prompts
Search the database for similar prompts to use as inspiration:
```bash
python3 {PROMPT_ENGINE_DIR}/scripts/search_prompts.py "SUBJECT" --model MODEL --limit 5
```
### Step 3: Construct the Prompt
Use this universal structure (adapt per model):
**Image Prompts:**
```
[Subject/Action] + [Setting/Environment] + [Lighting/Mood] + [Camera/Lens] +
[Style Reference] + [Technical Parameters]
```
**Video Prompts:**
```
[Camera Movement] + [Subject/Action] + [Setting] + [Lighting] +
[Duration/Speed] + [Style] + [Technical]
```
**Text Prompts:**
```
[Role/Context] + [Task] + [Constraints] + [Format] + [Tone] + [Examples]
```
### Step 4: Apply Model-Specific Syntax
Load `{PROMPT_ENGINE_DIR}/references/model-guide.md` for model-specific formatting rules:
- **Midjourney**: Use `--ar`, `--v`, `--style`, `--s`, `--chaos`
- **Flux**: Natural language, detailed descriptions work best
- **DALL-E**: Direct, descriptive, avoid negative prompts
- **Leonardo AI**: Use model-specific keywords (Phoenix, Alchemy)
- **Sora**: Camera movements, scene transitions, temporal descriptions
### Step 5: Review and Refine
Present the prompt with:
1. The full prompt text
2. Recommended parameters/settings
3. Similar prompts from the database for comparison
4. Suggested variations (3 alternatives)
## Quality Gates
- Prompt must be > 30 characters
- Must include subject + at least one modifier (style, lighting, or camera)
- Must use model-appropriate syntax
- No conflicting instructions (e.g., "realistic" + "cartoon")
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.