convez.ai
SkillIntegração & APIsAvançado
tr0tsky0

SD.AI Claude Skills

tr0tsky0/sdai-claude-skills

Pipelines de engenharia de prompts para criação de personagens e imagens em IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Conjunto de skills para Claude que implementa pipelines de engenharia de prompts especializados em criação de personagens e geração de imagens para plataformas de IA. Inclui metodologias testadas em campo para construção de perfis, arquitetura de mundos reativos e otimização de prompts multi-engine.

Instalação

$npx convez-ai tr0tsky0/sdai-claude-skills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Gerar personagens completos com perfis, ganchos narrativos e vinhetas iniciais
  • Otimizar prompts de imagem para múltiplos engines com tiers de conteúdo
  • Criar mundos reativos com personagens secundários e arquitetura de notas estruturada
  • Engenharia de prompts para vídeos de animação e cosplay

Veredito

Skills especializados e bem documentados para um caso de uso específico (plataforma adulta de IA). Implementação técnica sólida com pipeline modular, mas público-alvo restrito e dependência direta de plataforma proprietária limitam aplicabilidade geral.

README

A tradução para o português deste README ainda está na fila.

SD.AI Claude Skills

Two Claude skills for Secret Desires AI (SD.AI), an adult AI companion platform. These skills encode field-tested methodology for character creation and image prompt generation into reusable, invocable skill documents for Claude.ai.

Contents

SkillWhat it does
sd-character-genFull character creation pipeline: narrative hooks, character profiles, profile picture prompts, opening vignettes, and World Character creation with reactive world architecture
sd-image-promptsImage prompt engineering for SD.AI's four engines (DaVinci, Vermeer, Monet, Picasso), including engine selection logic, NSFW tier prompting, cosplay prompting, and Live Photo video generation

Requirements

Adult Content Notice

These skills are designed for use on Secret Desires AI, an age-gated adult AI companion platform. SD.AI requires age verification; its users have explicitly consented to adult content on that platform. The skills contain explicit anatomical terminology required for accurate image generation on SD.AI's adult content tiers — this is an engine requirement, not a stylistic choice. They are intended for use by adults with active SD.AI accounts.

Installation

  1. Download the .skill file for the skill you want (sd-character-gen.skill or sd-image-prompts.skill)
  2. Go to claude.ai and navigate to Skills in the sidebar
  3. Click Add skill and upload the .skill file
  4. The skill will be available in any conversation via /sd-character-gen or /sd-image-prompts

Usage

sd-character-gen

Invoke with /sd-character-gen. Use when you want to:

  • Generate narrative hooks for a new character concept (Stage 1)
  • Build a complete character profile with all SD.AI platform fields (Stage 2)
  • Generate profile picture prompts at SFW / Suggestive / NSFW tiers (Stage 3)
  • Write an opening vignette to seed the chat (Stage 4)
  • Create a World Character — a character embedded in a reactive world with lore, secondary characters, and a full Notes field architecture

The skill handles the full pipeline. You can enter at any stage with a concept already in hand, or start from scratch and let Stage 1 generate scenario hooks to choose from.

sd-image-prompts

Invoke with /sd-image-prompts. Use when you want to:

  • Generate prompts for any SD.AI engine with engine selection guidance
  • Get three content tiers (SFW / Suggestive / NSFW) in a single pass
  • Create cosplay prompts with proper costume anchoring
  • Generate Live Photo (video) animation prompts with joint-by-joint motion coverage
  • Debug prompts that aren't producing the expected output

Repository Structure

sd-character-gen/
├── SKILL.md                         # Main skill document
└── references/
    ├── platform-structure.md        # All SD.AI platform fields and presets
    ├── output-template.md           # Stage 2 character profile output template
    └── llm-embodiment.md            # Writing principles for LLM persona copy
sd-image-prompts/
└── SKILL.md                         # Main skill document
sd-character-gen.skill               # Packaged skill — upload directly to Claude.ai
sd-image-prompts.skill               # Packaged skill — upload directly to Claude.ai

The .skill files are zip archives of each skill directory. To rebuild after editing:

# Windows PowerShell
# Note: Compress-Archive doesn't support .skill extension, and ZipFile.CreateFromDirectory
# stores backslash paths on Windows which Claude.ai rejects. Build entries manually instead.
Add-Type -Assembly System.IO.Compression
$root = (Get-Location).Path
foreach ($skill in @("sd-character-gen", "sd-image-prompts")) {
    $src = "$root\$skill"; $tmp = "$root\$skill.skill.tmp"; $dest = "$root\$skill.skill"
    $s = [System.IO.File]::Open($tmp, [System.IO.FileMode]::Create)
    $z = New-Object System.IO.Compression.ZipArchive($s, [System.IO.Compression.ZipArchiveMode]::Create, $false)
    Get-ChildItem -Path $src -Recurse -File | ForEach-Object {
        $rel = $_.FullName.Substring($src.Length + 1).Replace('\', '/')
        $e = $z.CreateEntry($rel, [System.IO.Compression.CompressionLevel]::Optimal)
        $es = $e.Open(); $b = [System.IO.File]::ReadAllBytes($_.FullName)
        $es.Write($b, 0, $b.Length); $es.Close()
    }
    $z.Dispose(); $s.Close()
    [System.IO.File]::Copy($tmp, $dest, $true); [System.IO.File]::Delete($tmp)
    Write-Output "Built $skill.skill"
}
# macOS / Linux
cd sd-character-gen && zip -r ../sd-character-gen.skill . && cd ..
cd sd-image-prompts && zip -r ../sd-image-prompts.skill . && cd ..

License

CC BY-NC-SA 4.0 — Attribution required, non-commercial use only, derivatives must use the same license.

Ver o original no GitHub

Bom para

Relacionados