higgsfield-stack

Coordinates prompt construction with Higgsfield CLI, MCP connector, and bundled skills execution surfaces.

Updated Jul 15, 2026
One-click install
npx skills add https://github.com/executiveusa/buffer-blaster- --skill higgsfield-stack-executiveusa
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: higgsfield-stack
Source: https://github.com/executiveusa/buffer-blaster-/tree/main/skills/higgsfield/skills/higgsfield-stack
Command: npx skills add https://github.com/executiveusa/buffer-blaster- --skill higgsfield-stack-executiveusa

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Users working with Higgsfield's official tooling (CLI, MCP connector, bundled skills) alongside a prompt-engineering skill often face confusion about which layer owns prompt construction versus execution, leading to duplicated logic, invalid CLI flags, and wasted generation credits. ## Core Features & Use Cases - Layer Separation Rules: Defines that prompt construction (MCSLA structure, vocabulary, negative constraints) stays in this skill while authentication, uploads, job submission, and polling belong to Higgsfield's CLI, MCP, or bundled skills. - Two-Step Preflight Discipline: Documents schema verification (higgsfield model get <model> or models_explore) followed by cost estimation (higgsfield generate cost or get_cost: true) before any paid generation. - Detection and Handoff Templates: Provides signals for detecting which execution surface is present and one-line handoff templates for CLI, MCP, or bundled skills. - Use Case: A user in Claude Code asks for a Kling 3.0 video prompt and mentions having the Higgsfield CLI installed; the skill produces the prompt, appends a preflight cost-check recommendation, and hands off with the correct higgsfield generate create invocation line. ## Quick Start Ask the assistant to write a Higgsfield video prompt and mention that you have the Higgsfield CLI or MCP connector installed so it can add the correct preflight and handoff guidance.

Frequently Asked Questions about higgsfield-stack

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check Higgsfield generation cost before submitting a job?

Run a two-step preflight: first verify the model's parameter schema with higgsfield model get <model>, then estimate cost with higgsfield generate cost <model> plus your flags. On the MCP connector, use models_explore followed by generate_image or generate_video with get_cost set to true.

Should I use the Higgsfield CLI or the MCP connector?

Use the CLI for Claude Code, Codex, headless, or CI workflows since it uses long-lived API tokens. Use the MCP connector for conversational generation inside claude.ai web or the Claude desktop app. Queue priority depends on your plan tier, not the surface.

What is the difference between the two higgsfield-soul skills?

Higgsfield's bundled higgsfield-soul trains a Soul Character identity model from 5-20 face photos and returns a reference_id. The prompt-side higgsfield-soul sub-skill constructs prompts that use an already-trained identity. Theirs trains; ours prompts.

Why does higgsfield account balance fail in the CLI?

account balance and account credits are not valid subcommands and fall through to parent help. The canonical command for balance, plan, and email is higgsfield account status, with the alias higgsfield acc status.

Can this skill call the Higgsfield API directly instead of the CLI?

No. Direct calls to api.higgsfield.ai are explicitly forbidden because authentication, uploads, retries, polling, and rate-limit handling live inside the official CLI. If no CLI is present, the user pastes the prompt into higgsfield.ai manually.

When should the Seedance lint preflight run before generation?

Run python3 scripts/seedance_lint.py on prompts targeting Seedance 2.0 or Seedance Pro before submission. The content filter rejects prompts before GPU execution but still charges credits, so catching rejection patterns at prompt-construction time avoids wasted spend.