higgsfield-ai-prompt-skill

Higgsfield AI Prompt Skill

A comprehensive Claude skill library for generating high-quality prompts on Higgsfield AI — the cinematic video and image generation platform.

What This Skill Does

Transforms natural language requests into production-ready Higgsfield prompts using:

• The MCSLA formula (Model · Camera · Subject · Look · Action)
• Named camera controls and motion presets the platform recognizes
• Model selection guidance across Kling 3.0 / 3.0 Omni / 3.0 Motion Control, Veo 3.1, Wan, Seedance 2.5 / 2.0, FLUX 3 Video, Minimax Hailuo, Higgsfield DoP, and more
• Genre recipe templates for action, horror, romance, sci-fi, product ads, and more
• Soul ID character consistency guidance + Character Sheet creation
• Troubleshooting for failed or poor generations
• Cinema Studio 2.5 advanced features: Soul Cast AI actors, built-in color grading, 3D Mode (Gaussian Splatting), Grid Generation, Resolution Settings, Frame Extraction Loop, Object & Person Insertion, Per-Character Emotions, Clustering, Five-View Location Reference Sheet, Reference Sheet Types (Motion / Outfit / Palette), Elements System with library surface (5 source tabs × 6 element categories)
• Cinema Studio 3.0 (Business/Team plan): native dual-channel stereo audio, Smart shot control, 15s max duration, 7 genres, @ reference patterns, Soul Cast 3.0
• Cinema Studio 4.0 (cinematic_studio_video_4_0, CLI workflow): the Seedance 2.5 four-mode surface (t2v / omni_reference / video_edit / video_extension) plus camera body / lens / aperture / era / genre / pacing and light / palette controls — schema-documented, not yet field-rated
• 3D (higgsfield-3d): image / multi-view / text → GLB meshes, rigging + a 678-clip animation library, remesh / retexture, 3D Jutsu scene projects, 3D turnarounds and staging blockouts
• Repurposing (higgsfield-repurpose): Shorts Studio, Clipify, Virality Predictor, Video Analysis — with the paid / free split
• Cinema Studio 3.5: three-pill main UI (Genre / Style / Camera), Style Settings panel (8 Color Palette / 6 Lighting / 9 Camera Moveset Style + Manual Style mode), Camera Settings four-axis panel (3 Camera Body / 5 Lens / 5 Focal Length including new 75mm / 3 Aperture), Image Mode with four Cinematic models picker (Soul Cinema default, Cinematic Characters, Cinematic Locations, Cinematic Cameras with 2.5 vocabulary)
• Seedance 2.0 prompting best practices — Intent over Precision, Genre Router, I2V Gate, Anti-Slop, Physics Language, SCELA audio, Reference-Based / Continuation / Expand Shot / Edit Shot / Transformation prompt modes, Continuation Prompt Formula, the Iteration Rule
• Seedance 2.5 omni-reference director — the four generation modes (t2v / omni_reference / video_edit / video_extension), explicit @Image / @Video / @Audio reference roles with exclusions, the five-step multi-reference workflow, 30-second staging with end states, timestamp pacing, bracket syntax for music / SFX / dialogue / subtitles, first-last-frame control (platform start_image/end_image in omni_reference, or declared in-prompt) and multi-keyframe control, storyboard grids, coarse-vs-fine blockout rendering, one-click video, seamless transitions, and a 2.0-vs-2.5 routing table
• Hell Grind feature-film pipeline — Higgsfield’s open-sourced 95-minute-feature system: headless character sheets, mask-composited point edits, location sheets with anchors and one light logic, the per-scene GEO SPATIAL LAYOUT block, the position-fixing first second, dialogue construction, the ban dictionary, the 10–15 iteration rule, and the crowd / giant / threshold solutions
• AI-VFX pipeline — the production layer under Seedance 2.5, from Higgsfield’s AI-vs-VFX challenge build: model routing per asset class (faces / creatures / clothing / locations), the plain-grey sheet law, two-close-up creature sheets, the face-lock crop, the size-ref frame for scale between two subjects, location batching selected on light, the omni_reference video-to-video lane (source ≥4s, duration = source) with its performance-inheritance clause, the four-batch stop rule and the i2v fallback with a deliberate empty-frame stitch point, and a slop catalog of the tells that give a shot away
• Acting director — performance as behavior under pressure: objective / obstacle / tactics / beats / subtext, listening markers, body + status + proxemics, mandatory eye life, the 150–220-word acting master profile and its per-scene rewrite, the locked voice prompt, a 15-symptom atlas of bad acting, and a 0–5 performance scale
• GPT Image 2.0 prompt director — three-format taxonomy (structured JSON for UI mockups / infographics / reference sheets, dense cinematic prose for single-subject scenes, auto-derive meta-prompt for theme-only concepts) plus reference-sheet and static-ad-recreation workflow satellites
• Higgsfield Canvas — node-based / infinite-board workspace guidance: chaining prompts → images → videos, named canvas patterns, Shared Canvas live collaboration, build-free / generate-paid cost model
• Marketing Studio + Content Factory — 9 DTC ad presets (UGC / Tutorial / Unboxing / Hyper Motion / TV Spot / Wild Card / Virtual Try-On) with 4–15s ad video, plus an end-to-end campaign pipeline (research → plan → generate → publish → report) with a cost-savings report
• Shared negative constraints reference — categorized artifacts + prevention phrases (positive alternatives for 3.0); Kling 3.0 Motion Control failure diagnostic; Physics Rendering — Resolution Decision Matrix (cross-model 480p / 720p / 1080p routing rule for Seedance 2.0 + Cinema Studio 3.x)
• Identity vs. Motion separation — hard rule for character consistency across shots
• Annotated templates library — 10 genre templates with Cinema Studio 3.0 genre mappings, plus Seedance technique + character-design + text-overlay sub-libraries (19 files across 3 categories)
• DISCIPLINE.md cross-cutting framework — 9 named discipline patterns in 3-3-3 tier symmetry (prompt-construction, model-selection, iteration-discipline) governing decisions across all sub-skills
• production-benchmarks.md — Hell Grind 90-min Cannes feature reference, per-character iteration anchors, acceptance-rate calibration; what “production quality” means in practice
• FAILURE-MODES.md (Seedance) — 15 named render failures documented with symptom + mechanism + counter for diagnosis-first iteration
• C-arc Building Complete AI Projects — 10-Step Methodology — end-to-end pipeline from idea to delivered project; complements the genre/scene templates
• Expanded Seedance methodology + Soul refinement — Iteration Rule + 6-Pass Diagnostic Sequence + Four Questions + Next-Shot Decision Tree + Bridging / Continuation / Repair working modes; Character Anchor Block + Two-Tool Refinement Pipeline for character consistency at production scale

Install

Claude Code

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

Claude Code registers the root SKILL.md as one skill, higgsfield. The 35 sub-skills under skills/ are not registered separately; the root routes to each by its full path and Claude opens it from there. The skill’s scripts need only python3 (standard library).

Your documents and the skill’s deliverables go in a workspace/ folder (input/, output/, processed/) inside the project you are working in, created on first use. Your documents and deliverables are never written into ~/.claude/skills/higgsfield/, so updating or reinstalling the skill never touches them. The learning memory and generation ledger live in ~/.higgsfield-skill/db/, outside the skill folder, so an update or reinstall keeps them too. The skill creates that folder on first use from its own db/, which also carries over the memory from an install older than v3.42.0. If you had two older installs (say a personal one and one inside a project), the second one’s own entries are kept in ~/.higgsfield-skill/unmerged/, not merged into the first; where lists them. In Cowork, or when you run the skill from a clone of this repo, the memory stays in the skill’s own db/. To see which folder is in use, run python3 scripts/higgsfield_memory.py where from the skill’s folder.

Claude Cowork

Drop the repo folder into your Cowork workspace. The skill dispatcher is at SKILL.md in the repo root.

Claude.ai Projects

Upload SKILL.md (root) as your project instruction base. Upload files from skills/ as project documents.

Higgsfield Stack Integration

This skill is the prompt-construction layer. Higgsfield ships official execution tooling — a CLI, an MCP custom connector, and a bundled skills package. They complement each other: this skill produces the prompt, their tooling executes it. None of their tooling is required for this skill to work — you can always paste prompts directly into higgsfield.ai. But if you want an end-to-end loop, you’ll want one of the three.

A Higgsfield account is required for any of the tooling below. Sign up at higgsfield.ai.

Higgsfield CLI

Command-line tool for terminal-native agents (Claude Code, Codex, Cursor). Per Higgsfield’s own guidance, prefer the CLI over the MCP if you’re working in a terminal.

• Repo: github.com/higgsfield-ai/cli
• Install: curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh or brew install higgsfield-ai/tap/higgsfield
• Auth: higgsfield auth login

Higgsfield MCP

Custom connector for claude.ai web and the Claude desktop app. Separate product from the CLI.

• Connector URL: https://mcp.higgsfield.ai/mcp
• Install: claude.ai → Settings → Connectors → Add custom connector → paste the URL above → sign in

Higgsfield Bundled Skills

Markdown skill bundle for agents that consume Cowork-style skills. All three skills drive the CLI under the hood.

• Repo: github.com/higgsfield-ai/skills
• Install: npx skills add higgsfield-ai/skills
• Skills included: higgsfield-generate, higgsfield-soul, higgsfield-product-photoshoot
• Invoke: /higgsfield:generate, /higgsfield:soul, /higgsfield:product-photoshoot

End-to-end example

How the layers fit together for a real request:

USER:    "Make me a cinematic chase scene through a night market.
          Use my trained Soul character — reference_id abc123."
   ↓
THIS SKILL — higgsfield-ai-prompt-skill
   • routes to higgsfield-prompt + higgsfield-camera + higgsfield-soul
   • picks Kling 3.0 (character-focused, supports --soul-id)
   • applies MCSLA: model, camera preset, subject, look, action
   • appends shared negative constraints
   • outputs a production-grade Higgsfield prompt
   ↓
PRE-FLIGHT (optional, recommended for Veo / Kling / Seedance video):

   SCHEMA VERIFY (recommended for any model you haven't called recently):
   CLI path:        higgsfield model get kling3_0
                    → returns schema: aspect_ratio enum, duration range,
                      mode/sound options, media roles
   MCP path:        models_explore(action="get", model_id="kling3_0")
                    → returns same schema as CLI

   COST ESTIMATE (no job submitted):
   MCP path:        generate_video(..., get_cost: true)
                    → returns credit cost + adjustments block

   CLI path:        higgsfield generate cost kling3_0 \
                      --prompt "<prompt from this skill>" \
                      --aspect_ratio 16:9 \
                      --duration 8
                    # (add reference flags as needed: --soul-id, --start-image,
                    #  --end-image — consult `higgsfield model get kling3_0`
                    #  for supported media roles)

   Bundled skills:  drop to CLI for the cost check (same auth, same workspace),
                    then invoke /higgsfield:generate

   Optional account checks (same data across surfaces):
   MCP path:        balance / transactions tools
   CLI path:        higgsfield account status
                    higgsfield account transactions --size 50

   Note: 2.35:1 is anamorphic STYLE vocabulary, not a valid Kling 3.0 output
         ratio. Output ratios are platform-bounded: 16:9 / 9:16 / 1:1 only.
   ↓
HIGGSFIELD STACK — one of three execution surfaces:

   CLI path:
     higgsfield generate create kling3_0 \
       --prompt "<prompt from this skill>" \
       --aspect_ratio 16:9 \
       --duration 8 \
       --wait
     # (add reference flags as needed: --soul-id, --start-image, --end-image —
     #  consult `higgsfield model get kling3_0` for supported media roles)

   Bundled skills path:
     /higgsfield:generate — takes the prompt as its --prompt argument,
     formats the CLI call above under the hood

   MCP path (claude.ai web/desktop):
     Claude invokes the Higgsfield connector with the prompt as input
   ↓
USER:    Result URL returned. Iterate if needed (this skill's
         iteration discipline applies regardless of execution surface).

The layer split holds in every case: this skill always produces the prompt, the Higgsfield stack always handles the generation call. None of the three execution paths reach back into prompt construction; this skill never shells out to their CLI or API.

Full preflight discipline — when to surface it, marketing-studio caveat, CLI naming gotchas (account status, not balance), and the plan-tier-vs-surface framing — lives in skills/higgsfield-stack/SKILL.md § Preflight discipline.

Coexistence rules

For the full coexistence rules, detection signals, naming-collision callouts, and handoff templates, see skills/higgsfield-stack/SKILL.md.

Structure

.
├── SKILL.md                          ← Main dispatcher (routes to sub-skills — start here)
├── README.md                         ← This file
├── CHANGELOG.md                      ← Version history
├── CONTRIBUTING.md                   ← Contribution guidelines
├── LICENSE                           ← MIT license
├── CLAUDE.md                         ← Project instructions for Claude Code
├── .markdownlint.json                ← Linter config (CHANGELOG convention silencing — v3.6.1)
├── model-guide.md                    ← Model comparison tables + decision flowchart
├── image-models.md                   ← Image model reference + pricing tiers
├── vocab.md                          ← Full platform vocabulary reference
├── prompt-examples.md                ← High-quality example prompts + Before/After pairs
├── photodump-presets.md              ← Photodump mode presets
├── DISCIPLINE.md                     ← Cross-cutting discipline framework (9 patterns, 3-3-3 tier symmetry)
├── production-benchmarks.md          ← Production-quality anchors + acceptance-rate calibration
├── scripts/                          ← Python tooling (run from the repo root)
│   ├── higgsfield_memory.py          ← Memory system script
│   ├── seedance_lint.py              ← Seedance preflight linter
│   ├── validate.py                   ← Pre-release validation script
│   ├── build_index.py                ← Regenerates INDEX.md + checks QUICK FACTS anchors
│   ├── sync_specs.py                 ← Regenerates specs/ from a models_explore snapshot
│   ├── refresh_specs.py              ← Spec-drift tripwire (live CLI vs baseline; params, ranges, media roles)
│   ├── snapshot_crosscheck.py        ← Two-way structural check: models_explore snapshot vs live CLI
│   ├── preflight.py                  ← Free any-model platform-constraint preflight (enums + CEL rules, fails closed)
│   ├── claims_lint.py                ← Doctrine claims vs today's specs (registry: evals/spec-claims.json)
│   ├── repo_walk.py                  ← Repo file walks that skip agent worktrees
│   ├── user_guide_content.py         ← USER-GUIDE content derived from disk (dependency-free)
│   ├── generate_user_guide.py        ← USER-GUIDE.pdf generator (Path B refactor — v3.7.0)
│   ├── validate_user_guide.py        ← USER-GUIDE.pdf drift validator (text-extract + binary diff)
│   └── sub_skill_descriptions.py     ← Canonical sub-skill roster (shared data module)
├── db/
│   ├── filter-memory.json            ← Content filter memory (seeded)
│   └── quality-memory.json           ← Quality failure memory (seeded)
├── docs/                             ← Extended reference documents
│   ├── Seedance 2 Skill.md           ← Bilingual EN+ZH Seedance director reference
│   ├── archive/                      ← Historical records
│   │   ├── HISTORY.md                ← Consolidated v3.0.0–v3.6.0 audit + inventory snapshots
│   │   └── AUDIT-2026-06-03.md       ← Full repo audit (security, bugs, docs hygiene)
│   └── user-guide/                   ← Exported USER-GUIDE.pdf + current-version baseline (rotate, not accumulate)
├── templates/                        ← Genre templates + Seedance coordination + text-overlays
│   ├── 01-cinematic-action-chase.md
│   ├── 02-product-ugc-showcase.md
│   ├── 03-horror-atmosphere.md
│   ├── 04-fashion-editorial.md
│   ├── 05-sci-fi-vfx.md
│   ├── 06-portrait-character-intro.md
│   ├── 07-landscape-establishing-shot.md
│   ├── 08-comedy-social-media.md
│   ├── 09-romantic-intimate.md
│   ├── 10-dance-music-performance.md
│   ├── seedance/                     ← Seedance technique templates (10)
│   │   ├── multi-character-anchor.md
│   │   ├── single-character-position.md
│   │   ├── top-down-map.md
│   │   ├── omni-reference-2-5.md     ← Seedance 2.5 multi-reference brief
│   │   └── worked-example-two-character.md
│   └── text-overlays/                ← Text overlay templates
│       ├── slogan.md
│       ├── speech-bubble.md
│       └── subtitle.md
└── skills/
    ├── shared/
    │   └── negative-constraints.md       ← Shared artifact prevention reference
    ├── higgsfield-prompt/SKILL.md        ← Core MCSLA formula + prompt structure + Identity/Motion separation
    ├── higgsfield-image-shots/SKILL.md   ← Cinematic image prompting (shots, angles, composition)
    ├── higgsfield-gpt-image-2/
    │   ├── SKILL.md                      ← GPT Image 2.0 director (JSON / prose / meta-prompt taxonomy)
    │   ├── reference-sheet-workflow.md   ← Automatic product reference-sheet workflow
    │   └── static-ads-workflow.md        ← Static-ad recreation workflow
    ├── higgsfield-models/
    │   ├── SKILL.md                      ← Compact model selection guide
    │   └── MODELS-DEEP-REFERENCE.md      ← Full per-model documentation (on-demand)
    ├── higgsfield-camera/SKILL.md        ← All camera controls + usage
    ├── higgsfield-motion/SKILL.md        ← Named motion presets library
    ├── higgsfield-style/SKILL.md         ← Visual styles + color grades + lighting
    ├── higgsfield-soul/SKILL.md          ← Soul ID character consistency
    ├── higgsfield-audio/SKILL.md         ← Audio prompting + Cinema Studio 3.0 native audio
    ├── higgsfield-apps/SKILL.md          ← One-click Apps guide
    ├── higgsfield-recipes/SKILL.md       ← Genre scene templates
    ├── higgsfield-troubleshoot/SKILL.md  ← Fix failing generations
    ├── higgsfield-assist/SKILL.md        ← General assistant + platform guidance
    ├── higgsfield-mixed-media/SKILL.md   ← Mixed media + hybrid generation
    ├── higgsfield-moodboard/SKILL.md     ← Moodboard creation workflows
    ├── higgsfield-pipeline/SKILL.md      ← Multi-step generation pipelines
    ├── higgsfield-canvas/SKILL.md        ← Node-based Canvas workspace + named patterns + Shared Canvas
    ├── higgsfield-content-factory/
    │   ├── SKILL.md                      ← Campaign pipeline (research → plan → generate → publish → report)
    │   └── publish-and-report-workflow.md ← Publish + cost-savings report satellite
    ├── higgsfield-marketing-studio/
    │   ├── SKILL.md                      ← Marketing Studio: 9 ad presets + 4–15s ad video
    │   └── cross-surface-workflow.md     ← ms_image / DTC Ads cross-surface workflow
    ├── higgsfield-recall/SKILL.md        ← Recall + regeneration patterns
    ├── higgsfield-3d/SKILL.md            ← 3D meshes (image / multi-view / text → GLB), rigging, 3D Jutsu scenes
    ├── higgsfield-repurpose/SKILL.md     ← Finished video in: Shorts Studio, Clipify, Virality Predictor, Video Analysis
    ├── higgsfield-cinema/SKILL.md        ← Cinema Studio 2.5 + 3.0 + 3.5 + 4.0 (Soul Cast, Color Grading, 3D Mode, Smart Mode, @ References, Native Audio, three-pill UI, Image Mode, Cinematic models picker)
    ├── higgsfield-seedance/
    │   ├── SKILL.md                      ← Seedance 2.0 prompt director + content-filter preflight
    │   ├── ENGINE-RULES.md               ← Hard rendering constraints shared across the Seedance family
    │   ├── PRODUCTION-PATTERNS.md        ← Tutorial-demonstrated production patterns
    │   ├── HELL-GRIND.md                 ← Higgsfield's open-sourced 95-min feature pipeline
    │   └── FAILURE-MODES.md              ← 15 named Seedance render failures (symptom · mechanism · counter)
    ├── higgsfield-seedance-2-5/
    │   ├── SKILL.md                      ← Seedance 2.5 omni-reference dialect + 4-mode router
    │   ├── MODE-PLAYBOOKS.md             ← Edit / extend / storyboard / blockout / one-click / transitions
    │   └── VFX-PIPELINE.md               ← AI-VFX pipeline: asset routing, size-ref frame, omni_reference v2v, slop catalog
    ├── higgsfield-acting/SKILL.md        ← Performance craft: objective, beats, eye life, master profile
    ├── higgsfield-scene-engine/SKILL.md  ← Structural scene audit: goal · obstacle · tactic · reversal · value shift
    ├── higgsfield-vibe-motion/SKILL.md   ← Vibe-based motion direction
    └── higgsfield-workspaces/SKILL.md    ← Workspace-first decision layer (Cinema Studio / Lipsync / Draw-to-Video / Viral Hub presets / Click to Ad / Higgsfield Audio)

Generation Ledger

Every generation attempt — kept, rejected, or filter-flagged — gets one row in ledger/<project>.json in the memory folder, logged by the agent in ≤5 seconds (one question, one command — never a form). After ~30–40 rows a production has empirical takes-per-kept ratios per shot type instead of vibes:

python3 scripts/higgsfield_memory.py log-gen adze --model seedance_2_0 \
  --tags dialogue-cu,two-char --outcome rejected --reason extra-cuts
python3 scripts/higgsfield_memory.py ratio adze --credits     # hit rates + money view
python3 scripts/higgsfield_memory.py budget adze --shots plan.json   # price before burning

Tags and reject reasons come from controlled vocabularies (db/ledger/README.md); rows are append-only with superseding corrections; ratio splits structural vs stochastic rejections and flags low-n rows instead of faking precision.

Maintenance — two activation steps

Two capabilities ship inactive on purpose and need a one-time setup.

1. Activate the scheduled spec-drift check

.github/workflows/spec-drift.yml runs scripts/refresh_specs.py weekly to catch when Higgsfield changes a model’s lineup or capabilities before the 30-day staleness warning would. It ships dormant until you give it the Higgsfield CLI credentials as a repo secret:

higgsfield auth login                 # if not already authenticated locally
gh secret set HIGGSFIELD_CREDENTIALS < ~/.config/higgsfield/credentials.json

Then run it once manually (Actions tab → spec-drift → Run workflow) to confirm the CLI-install step resolves on the runner. After that it’s automatic: fresh → nothing; drift → it opens/updates a GitHub issue with next steps; auth expired → the job fails so GitHub notifies you to re-run higgsfield auth login and refresh the secret. The credentials live only in the GitHub secret — they are never committed.

2. Let routing telemetry accumulate before pruning

log-route / routing (in scripts/higgsfield_memory.py) record which sub-skills each request opens, so “which skills are load-bearing, which to retire” becomes a data question:

python3 scripts/higgsfield_memory.py log-route --skills higgsfield-prompt,higgsfield-camera
python3 scripts/higgsfield_memory.py routing     # ranks opens, lists the never-opened tail

This is instrumentation, not a verdict — let real requests accumulate before acting on the tail. A small sample is not evidence a skill is dead.

Example Prompts

Basic:

“Write me a Higgsfield prompt for a cinematic action chase through a night market”

Specific:

“I need a horror prompt using VHS style, Dutch angle camera, and the Horror Face preset”

With reference:

“I have a Soul ID character. Write 3 different scene prompts with her — office, party, rooftop”

Model question:

“Should I use Kling 3.0 or Seedance 2.0 for a large-scale battle scene?”

Troubleshoot:

“My image-to-video isn’t animating, it’s just static. What am I doing wrong?”

The MCSLA Formula

Acknowledgements

• @vitaesomniastore-commits — skill-authoring review of v3.40.0 against Anthropic’s skill best practices (#110), which shaped v3.41.0.

Built February 2026 · v3.42.0 (updated 2026-10-23) · Platform: higgsfield.ai