| name | ai-artist |
| description | Prompt engineering for AI image generation: Midjourney, DALL-E, Stable Diffusion, Flux. Use when writing image prompts, configuring generation parameters, or optimizing AI art output. NOT for UI design decisions (use design-system) or media file processing (use media-processing). |
| metadata | {"author":"pikakit","version":"3.9.206","category":"prompt-engineering","triggers":["prompt","AI prompt","image generation","Midjourney","DALL-E","Stable Diffusion","Flux"],"coordinates_with":["studio","media-processing","context-engineering"],"success_metrics":["100% deterministic prompt reproduction"]} |
AI Artist — Prompt Engineering
Deterministic prompt construction for LLMs and image generation models. Same inputs = same output.
Prerequisites
Required: None — AI Artist is a knowledge-based skill with no external dependencies.
Optional:
- Target model API access (Claude, GPT, Gemini, Midjourney, DALL-E, SD, Flux)
studio skill (for design system prompts)media-processing skill (for post-processing generated images)
When to Use
| Situation | Reference |
|---|
| Structuring LLM prompts | This file — LLM Prompt Pattern |
| Generating image prompts | rules/image-prompts.md |
| Marketing/copywriting prompts | rules/domain-marketing.md |
| Code generation prompts | rules/domain-code.md |
| Model-specific parameters | rules/model-syntax.md |
| Architecture review, contracts, security | rules/engineering-spec.md |
System Boundaries
| Owned by This Skill | NOT Owned |
|---|
| Prompt template construction | API submission to models |
| Domain-specific prompt patterns | Response quality evaluation |
| Model parameter syntax | Content safety filtering |
| Success criteria extraction | Token counting (→ context-engineering) |
| Prompt versioning structure | Design systems (→ studio) |
Pure function: This skill produces text output. Zero network calls, zero file mutations, zero side effects.
Execution Model — 3-Phase Lifecycle
| Phase | Action | Output |
|---|
| Parse | Validate domain, template, parameters | Validated input or error |
| Compose | Apply template, inject parameters, format for model | Structured prompt string |
| Emit | Return prompt with metadata and success criteria | Complete output |
All phases execute synchronously. No async pipeline, no queue.
LLM Prompt Pattern
[Role] You are a [expertise] specializing in [domain].
[Context] [Background, constraints, audience]
[Task] [Specific action to perform]
[Format] [Output structure: markdown, JSON, etc.]
[Constraints] [Word limit, tone, technical level]
[Examples]
Good: [Example of desired output]
Bad: [Example to avoid]
Image Prompt Pattern
[Subject] + [Style] + [Composition] + [Quality] + [Model Parameters]
Parameter ordering is fixed: subject → style → composition → quality → model_params.
Example:
Portrait of cyberpunk hacker, neon purple lighting, cinematic close-up,
volumetric fog, 8k render, artstation quality --ar 16:9 --style raw
| Element | Purpose | Example |
|---|
| Subject | What to generate | "portrait of cyberpunk hacker" |
| Style | Visual aesthetic | "neon lighting, cinematic" |
| Composition | Framing | "close-up, shallow depth of field" |
| Quality | Render markers | "8k, artstation quality" |
| Parameters | Model-specific | --ar 16:9 --style raw |
Supported Models
| Model | Domain | Parameter Reference |
|---|
| Claude, GPT, Gemini | Text, Code, Marketing | Core LLM pattern (above) |
| Midjourney | Image | rules/model-syntax.md — --ar, --style, --chaos |
| DALL-E | Image | rules/model-syntax.md — size, quality, style |
| Stable Diffusion | Image | rules/model-syntax.md — steps, cfg, sampler |
| Flux | Image | rules/model-syntax.md — guidance, steps |
Unknown model → generic format (no model-specific parameters injected).
Error Taxonomy
| Code | Recoverable | Trigger |
|---|
ERR_INVALID_DOMAIN | No | Domain not one of: text, image, code, marketing |
ERR_INVALID_TEMPLATE | No | Template not one of: role-task, subject-style, chain-of-thought, few-shot |
ERR_MISSING_PARAM | Yes | Required parameter is null or empty |
ERR_UNKNOWN_MODEL | Yes | Model not recognized; falls back to generic format |
ERR_TEMPLATE_NOT_FOUND | No | Reference file missing from skill directory |
ERR_PARAM_CONFLICT | Yes | Contradictory parameters (e.g., format=json + template=subject-style) |
ERR_PARAM_TOO_LONG | Yes | Parameter exceeds maximum character limit |
Zero internal retries. Deterministic output makes retries meaningless for same input.
Anti-Patterns
| ❌ Don't | ✅ Do |
|---|
| "Make it better" | "Increase contrast by 20%" |
| "Professional but casual" | Choose one tone, specify explicitly |
| Skip context/constraints | Set role, audience, constraints upfront |
| Over-prompt with 10+ directives | Focus on 3–5 core requirements |
| Assume model compatibility | Check rules/model-syntax.md for target model |
Troubleshooting
| Problem | Cause | Resolution |
|---|
| Vague output | Missing constraints | Add specific criteria: "under 150 words", "technical audience" |
| Wrong visual style | Style description too generic | Use style references: "in the style of [artist/work]" |
| Busy image composition | Too many subjects | Reduce to one primary subject, simplify background |
| Inconsistent results | No seed parameter | Add --seed for Midjourney/SD; not controllable for LLMs |
| Syntax errors in model params | Wrong model format | Check rules/model-syntax.md for correct syntax |
| Token limit exceeded | Prompt too long | Estimate: ~4 chars/token for English; trim constraints |
📑 Content Map
🔗 Related
| Item | Type | Purpose |
|---|
studio | Skill | Design system generation, color palettes |
media-processing | Skill | Post-processing generated images |
context-engineering | Skill | Token management, context window planning |
⚡ PikaKit v3.9.206
