/ Directory / Playground / higgsfield-ai-prompt-skill
● Community OSideMedia ⚡ Instant

higgsfield-ai-prompt-skill

by OSideMedia · OSideMedia/higgsfield-ai-prompt-skill

18 sub-skills to craft cinematic Higgsfield AI prompts using the MCSLA formula, Soul ID, and 10 genre templates — pro video-gen prompting as a skill.

A skill bundle for Higgsfield AI prompt engineering. Wraps the MCSLA formula (Movement, Character, Setting, Lighting, Atmosphere), Soul ID character consistency, and Cinema Studio 2.5 settings into prompt templates keyed to specific genres — noir, action, dreamcore, docu, etc.

Why use it

Key features

Live Demo

What it looks like in practice

higgsfield-ai-prompt-skill.replay ▶ ready
0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "higgsfield-ai-prompt-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
        "~/.claude/skills/higgsfield-ai-prompt-skill"
      ],
      "_inferred": true
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "higgsfield-ai-prompt-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
        "~/.claude/skills/higgsfield-ai-prompt-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "higgsfield-ai-prompt-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
        "~/.claude/skills/higgsfield-ai-prompt-skill"
      ],
      "_inferred": true
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "higgsfield-ai-prompt-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
        "~/.claude/skills/higgsfield-ai-prompt-skill"
      ],
      "_inferred": true
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "higgsfield-ai-prompt-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
        "~/.claude/skills/higgsfield-ai-prompt-skill"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json
{
  "context_servers": {
    "higgsfield-ai-prompt-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/OSideMedia/higgsfield-ai-prompt-skill",
          "~/.claude/skills/higgsfield-ai-prompt-skill"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add higgsfield-ai-prompt-skill -- git clone https://github.com/OSideMedia/higgsfield-ai-prompt-skill ~/.claude/skills/higgsfield-ai-prompt-skill

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use higgsfield-ai-prompt-skill

Produce your first genuinely cinematic Higgsfield clip

👤 Creators new to Higgsfield who keep getting generic-looking output ⏱ ~20 min beginner

When to use: Your test clips all look AI-flat and you want the richer look.

Prerequisites
  • Higgsfield account — higgsfield.ai — free tier available
  • Skill installed — git clone https://github.com/OSideMedia/higgsfield-ai-prompt-skill ~/.claude/skills/higgsfield-ai-prompt-skill
Flow
  1. Pick a genre
    Use higgsfield-ai-prompt-skill. I want a noir detective shot, rainy night, neon reflections. Give me an MCSLA-formatted prompt.✓ Copied
    → Structured prompt with all 5 MCSLA elements specified
  2. Refine lighting
    Make the lighting more Blade Runner 2049 — single harsh key, heavy atmospheric haze.✓ Copied
    → Lighting directive referencing specific cinematographic terms
  3. Add camera movement
    Camera: slow push-in, 35mm anamorphic feel.✓ Copied
    → Camera directive in Higgsfield's accepted syntax

Outcome: A prompt that actually produces the look you wanted.

Pitfalls
  • Too many adjectives — MCSLA rewards specificity, not density — one great adjective beats five mid ones

Keep a character consistent across shots with Soul ID

👤 Creators making multi-shot sequences with recurring characters ⏱ ~30 min intermediate

When to use: Shot 2 doesn't look like the same person as shot 1.

Flow
  1. Establish the character
    Use higgsfield-ai-prompt-skill. Build a Soul ID descriptor for my protagonist: mid-30s, asymmetric haircut, oval wire glasses, tan overcoat.✓ Copied
    → Reusable descriptor block
  2. Reuse across shots
    Now write 5 shots reusing this Soul ID, each in a different setting.✓ Copied
    → 5 prompts with the descriptor block embedded identically

Outcome: Shots where the character actually looks like the same person.

Pitfalls
  • Descriptor drifts shot-to-shot — Paste it verbatim; don't rewrite

Generate a music-video shot list with consistent visual grammar

👤 Indie musicians, visual creators ⏱ ~45 min intermediate

When to use: You want a coherent look across 20+ shots, not a chaotic mix.

Flow
  1. Set the visual grammar
    Use higgsfield-ai-prompt-skill with the music-video template. Visual grammar: dreamcore pastels, 16mm grain, slow dolly.✓ Copied
    → Visual grammar block you can prepend to every shot
  2. Generate shot list
    Now give me 15 shots mapped to the song's arc — intro, verses, chorus, bridge, outro.✓ Copied
    → Structured shot list with camera and subject per shot

Outcome: A shot list you can feed into Higgsfield in sequence.

Combinations

Pair with other MCPs for X10 leverage

higgsfield-ai-prompt-skill + seedance2-0-shotdesign-skill

Combine Higgsfield prompt craft with Seedance shot design for a cross-model workflow

Use Higgsfield MCSLA for the look and Seedance shot design for blocking.✓ Copied
higgsfield-ai-prompt-skill + filesystem

Save prompt library for reuse across projects

Save the Soul ID and genre blocks into ~/higgsfield-library/✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
mcsla_prompt scene description Every shot 0
soul_id_build character description First appearance of a character 0
genre_template genre name Kicking off a project 0
cinema_studio_tune desired look Final-pass parameter tuning 0

Cost & Limits

What this costs to run

API quota
Depends on Higgsfield plan
Tokens per call
Modest — prompts are text
Monetary
Higgsfield pricing separate; skill itself is free
Tip
Iterate on prompts locally before burning Higgsfield credits.

Security

Permissions, secrets, blast radius

Credential storage: No direct creds; Higgsfield usage happens outside skill
Data egress: None from skill

Troubleshooting

Common errors and fixes

Shots look nothing like the prompt

Higgsfield tuning matters — Cinema Studio 2.5 params are not optional

Character drifts between shots

Soul ID must be pasted identically; any wording change shifts the face

Alternatives

higgsfield-ai-prompt-skill vs others

AlternativeWhen to use it insteadTradeoff
Raw Higgsfield promptingYou're an experienced cinematographer and don't need formulasMore trial and error
seedance2-0-shotdesign-skillYou use Seedance instead of HiggsfieldDifferent model, different prompting quirks

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills