Generation skill from the official Higgsfield repository: The agent uses the higgsfield CLI to select models based on the task, generating images, videos, audio, and 3D assets, or scoring finished videos for "viral potential," returning only media links or text reports.
skills.sh install count 223276 (+1475 / 24h)
Security Audit: Not audited
- Not yet scanned
Rationale
- SKILL.md specifies default model selections: GPT Image 2.5 for general images and designs with text, Seedance 2.5 for video (the author states it supports 4–30 seconds, up to 1080p; use Seedance 2.0 for 4K), Nano Banana 2 for cartoon characters, Marketing Studio for ads, unboxing, and product showcase videos, and Seed Audio 1.0 for audio; other models are used only when explicitly requested by the user or when cheaper/faster options are needed
- One-step submission and wait: higgsfield generate create <model> --prompt ... --wait; reference images, first/last frames, reference videos, and audio are passed directly as local paths, with the CLI handling automatic uploads
- Virality Predictor (brain_activity) takes a finished video as input and returns scores for hook strength, attention, retention, etc., along with a report link; this is analysis, not generation
- skills.sh shows 1475 installs in 24 hours, with a cumulative total of 223276; the same repository contains 8 other skills (indicating only level of attention)
Usage
- Requires a Higgsfield account; if login expires, use higgsfield auth login; the author did not provide CLI installation instructions in the portion reviewed
- First check model IDs with higgsfield model list --json, then view parameters with higgsfield model get <model> --json; to check prices beforehand, use higgsfield generate cost
Risks
- Not audited in this period, fixed at C-level, and install commands are not displayed
- Generation consumes paid credits per use, while SKILL.md requires the agent to default to not estimating costs and to prioritize quality-first models; if cost is a concern, explicitly request a price quote before each generation in the prompt
- Your local images, videos, and face photos will be automatically uploaded to Higgsfield; visual quality has not been verified, and all stated model capabilities are as claimed by the author