What it does

female-portrait-director is a Codex Skill that takes sparse inputs—style, scene, outfit, mood—and expands them into complete, copy-ready prompts for AI image generators. It locks your explicit choices, fills in missing visual details (lighting, pose, camera framing, filters), and outputs negative constraints to keep generation stable. Think of it as a photography director who speaks fluent LLM.

The interesting bit

The architecture is deliberately modular: each request routes through exactly one on-demand style file, preventing keyword collisions that plague generic prompt libraries. The project also carries explicit safety boundaries—subjects must be clearly adult, with separate handling for e-commerce product shots versus “curve-focused” portraits—making the guardrails part of the engineering, not an afterthought.

Key highlights

• 14 supported styles, from “clean lifestyle” to “cold xianxia enhanced” to “Oriental voluptuous”
• Structured output: locked parameters → module analysis → final prompt → negative constraints
• One-command install via npx skills@latest add
• Reference-image workflows preserve authorized identity or product visuals
• Full multilingual docs (EN, ZH, JA, KO)

Caveats

• Requires OpenAI Codex; not usable standalone without the agent framework
• The “curve-focused” and “voluptuous” style categories suggest the safety boundaries exist for a reason—review docs/prompt_safety.md before production use
• No quantitative benchmarks or comparison data against raw prompting

Verdict

Worth a look if you’re generating female portraits at volume and tired of hand-crafting prompts or debugging style bleed. Skip it if you need male subjects, non-portrait genres, or a tool that works outside the Codex ecosystem.