본문으로 건너뛰기
cccskills
무료GitHub에 공개

higgsfield-content-factory

Use when the user wants to run a full ad-campaign pipeline on top of Higgsfield Marketing Studio — 'create a campaign', 'build a content plan', 'run the content pipeline', 'generate 100 UGC videos', 'plan and schedule a launch', 'make a batch of ads from my product', or 'how much did this save vs traditional production'. Covers the 5-stage orchestration (Research → Plan → Generate → Publish → Report), the UGC-first 5-format campaign mix (UGC Entertainment, Street Interview, Unboxing, Product Review, ASMR), the even-split allocation math, button-driven onboarding, the per-batch generation gate, and the publish + cost-report tail (satellite). Defers all Marketing Studio API ground-truth (presets, params, hooks, avatars) to higgsfield-marketing-studio.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

실행 점수 미검증

설치 방법 보기

포함된 파일(2)

  • SKILL.md14.8 KB
  • publish-and-report-workflow.md8.6 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Higgsfield Content Factory — Campaign Orchestration Pipeline

<!-- CONTENTS — generated by scripts/build_index.py; do not hand-edit -->

Contents

<!-- /CONTENTS -->

A 5-stage pipeline that runs a full ad campaign on top of Higgsfield Marketing Studio: Research → Plan → Generate → Publish → Report. Where higgsfield-marketing-studio is the engine reference (what the presets, parameters, hooks, and avatars are), Content Factory is the orchestration layer that drives a whole campaign turn by turn — researching trends, planning a dated content calendar, generating in batches, scheduling to ads, and reporting cost savings.

Translated from Adil Aliyev's higgsfield-content-factory source skill, per the v3.7.13 / v3.7.16 translation precedent. All Marketing Studio API ground-truth — preset slugs, parameter schema, hook/setting/avatar handling — lives in ../higgsfield-marketing-studio/SKILL.md and is NOT restated here. This sub-skill carries the pipeline scaffolding the marketing-studio sub-skill deliberately deferred. The publish + cost-report tail (Stages 4–5) lives in the companion publish-and-report-workflow.md satellite.


1. Two operating principles

Content is UGC-first. Every campaign defaults to a heavy share of UGC-family content — talking-head clips, honest reviews, ASMR close-ups, unboxing reveals. Hyper Motion, TV Spot, and Wild Card stay available but off the default mix. Every video idea must be producible inside one live Marketing Studio preset within its duration cap (see marketing-studio § 3 / § 7 for the canonical preset list and limits).

The interaction is button-driven. Every clarifying question is asked as a 2–4 option button choice (one round-trip), never a free-form "type your answer" for navigation, confirmation, or routing. Free-form typing is reserved for content the user must originate — and even then, always offer a smart default they can accept with one click. The only non-button steps are attaching a product image or pasting a product URL (both click-based file actions).

User-facing language rule (HARD). The user is not a developer. Do not narrate technical mechanics — no MCP tool names, no UUID resolution, no "running parallel searches," no enhanced-prompt internals. All of that runs silently. Send one plain-language stage banner when each stage starts, a brief "Stage N done — [deliverable]" between stages, then the next banner. The user sees a clean narrative arc, not a tool log. If the user asks "how does it work under the hood," then explain.

Stage banners

Send one banner per stage, in plain language (no tool names):

StageBanner
1🔍 Research & ideas — starting now. Scanning what's trending this week in your niche across Instagram, TikTok, and YouTube, then turning it into 15+ viral video ideas for your brand.
2🗂️ Content plan — starting now. Building your full video content plan as a polished document, every video mapped, dated, and ready to generate.
3🎬 Generating videos — starting now. Producing your videos in Marketing Studio, one preset at a time. I'll ask before each batch fires, so you stay in control.
3 (images)🖼️ Image asset pack — starting now. Generating your social posts, hero banners, and product stills via GPT Image 2.0.
4📅 Scheduling to ads — starting now. Setting up campaigns and scheduling across the calendar you approved.
5💰 Cost report — starting now. Compiling what you spent on Higgsfield versus what this volume would cost the traditional way.

2. Onboarding — one message, all buttons, no pauses

Ask the onboarding questions in a single button round-trip at the start. Do not ask them sequentially; do not pause between onboarding and the product request.

  • A — Higgsfield connection: "Yes — connected" · "Not yet — I'll connect now" · "Skip — research only."
  • B — Starting stage: Stage 1 full pipeline (needs a product) · Stage 2 build content plan (have a brief) · Stage 3 generate now (have a plan) · Stage 4 schedule (content ready). A product image with no other context defaults to Stage 1.
  • C — Video volume: 50 · 100 (recommended) · 150 · 200 · Other (any number). Store as [VIDEO_COUNT].
  • D — The product: "Attach your product image OR drop a URL — that's all I need to start." Skip D if a product is already attached. For Stage 3/4, swap D for "drop your existing content plan."

Send A + B + C as buttons and the product-attach prompt in the same message; once answered, proceed straight to the chosen stage with no extra confirmation.

Compute the per-format split silently the instant [VIDEO_COUNT] locks in — do not announce the breakdown here. It surfaces later, woven into the Stage 1 brief as a consequence of what the research showed, so it reads as a trend-driven choice rather than a config card.


3. The campaign mix — 5 UGC formats (even split)

Every campaign distributes evenly across five viral UGC formats. Cinematic secondary types (Hyper Motion / TV Spot / Wild Card) stay off-default — used only when the user explicitly asks.

#FormatShareWhat it is
1UGC Entertainment20%challenge / dare / entertainment-first; the product is the punchline
2Street Interview20%sidewalk stranger interviews where the product appears; high-trust "real people" feel
3Unboxing20%premium reveal energy — hands, packaging, the discovery moment
4Product Review20%honest talking-head; product in hand, label read aloud, ranking
5ASMR20%sound-led close-ups; caption-only, audible product handling

Allocation math. per_format = floor(VIDEO_COUNT / 5). If VIDEO_COUNT doesn't divide evenly, distribute the remainder one-per-format from format 1. Examples: 100 → 20/20/20/20/20; 50 → 10×5; 17 → 4/4/3/3/3.

This is the default mix and is user-overridable. The default share is a craft choice (UGC-first), not a platform rule.

Format → preset routing

Each format maps to a Marketing Studio preset. Route preset selection through the canonical mechanism documented in ../higgsfield-marketing-studio/SKILL.md § 3 (slug naming is canonical there; routing mode lives on a separate MCP call). At a glance:

  • Formats 1 (Entertainment), 2 (Street Interview), 5 (ASMR) → the UGC preset; what differentiates them is the system hook, the setting, the audio flag, and the prompt content — not a separate preset.
  • Format 3 (Unboxing) → the Unboxing preset.
  • Format 4 (Product Review) → the Product Review preset.

Do not hardcode hook/setting/avatar UUIDs or invent a generate_video.mode slug. Enumerate hooks/settings live (marketing-studio § 4 live-enumeration discipline) and pass the resolved IDs at generation time.

Concept seeds (vary within each format)

No two videos in the same format should be the same concept. Pull from seeds like: blind taste try, "$100 to try it" street challenge, will-it-pour (Entertainment); "rate it out of 10," "sing for the product," blind opinion → brand reveal (Street Interview); trio reveal, ribbon-pull solo drop, subscription-box drop (Unboxing); two-ingredient test, fridge-ranking, 7-day diary (Product Review); macro cap-unscrew + glug pour, condensation slide, spoon-clink ice-drop (ASMR).

ASMR is a style, not a preset

Render ASMR as the UGC preset with audio on, an intimate low-noise setting (Kitchen / Bathroom / Bedroom), usually no system hook, and a prompt focused on close-up handling sounds (cap unscrew, pour, glass clink, condensation). No talking; caption-only or silent open.


4. Stage 1 — Trend research & viral ideas

All research from live web searches; every idea validated against live Marketing Studio capability; UGC-first weighting.

  1. Probe capability silently. Pull the current preset / hook / setting picklists live (marketing-studio § 4 / § 5 enumeration). Cache the hook + setting IDs for Stage 3. Never narrate this.
  2. Auto-detect product & niche from the image or URL — do not ask the user to confirm.
  3. Run trend research — a spread of web searches across the product's niche on the major platforms, plus at least two source-page fetches, run silently.
  4. Synthesize a Viral Content Brief (UGC-first): 15+ idea cards, each tied to a format + preset, with the campaign's format breakdown woven in as a consequence of what the research surfaced.
  5. Producibility self-check before adding any idea — every idea must be renderable inside one preset within the duration cap. Ideas needing lip-sync from non-human characters, multi-character coordinated dialogue,

    15s single clips, or split-screen are out (or routed to an escape- hatch model — see § 6).

  6. Approve via buttons before moving to Stage 2.

5. Stage 2 — Video content plan

Turn the approved brief into a dated content plan: one row per video ([VIDEO_COUNT] rows), bucketed by the 5 formats, each with its format, preset, concept, hook/setting intent, and a calendar date. Render it as a polished, saved document deliverable the user can review and approve before any generation fires. Interleave dates so formats rotate across the calendar rather than clustering.

6. Stage 3 — Generate in Marketing Studio

The point of failure-avoidance here is the per-batch gate: never fire a whole campaign at once.

  1. Upload the product silently and register it (marketing-studio § 6).
  2. Resolve hooks / settings / avatars at runtime per the cached picklists; never hardcode IDs.
  3. Generate one preset batch at a time, and ask before each batch fires with a button gate ("Generate the 20 UGC Entertainment videos now?"). This keeps the user in control and prevents a runaway spend.
  4. No on-screen text in prompts — Marketing Studio renders text unreliably; keep copy out of the generation and add it in post or via the image pack.
  5. Image asset pack (optional) — generate social posts / hero banners / product stills via GPT Image 2.0, scaling the count from [VIDEO_COUNT] (see ../higgsfield-gpt-image-2/SKILL.md for prompt craft; the count split skews toward social posts, then hero, then with/without-people stills). A standalone image-pack mode skips the videos entirely.
  6. Handle failures by re-resolving the picklist and retrying the single failed batch, not the whole campaign.

Escape hatch — ideas Marketing Studio can't render

If an idea genuinely needs lip-sync, longer narrative, or multi-shot continuity, label it "Outside Marketing Studio" and route to Wan 2.7 (synchronized audio + character-consistent), Veo 3.1 (cinematic, audio- friendly), Cinema Studio Video 3.0 (cinema-grade), Seedance 2.0 (reference- driven, multi-SKU), or Kling 3.0 (multi-shot, motion transfer). Use sparingly — UGC-first means most ideas stay inside Marketing Studio.

7. Stages 4–5 — Publish & report

Stage 4 (schedule & publish to Meta Ads) and Stage 5 (cost-comparison report) live in the companion publish-and-report-workflow.md satellite. That tail carries the most connector-dependent and estimate-based material, so its caveats are isolated there.


8. Hard rules (consolidated)

  • Button-driven navigation; free-form typing only for user-originated content, always with a one-click default.
  • 5-format even split as the default mix; user-overridable.
  • Producibility check before any idea ships to the plan.
  • Per-batch generation gate — confirm before each batch fires.
  • Hooks are visual scene templates, not verbal copy; never confuse a hook with caption text.
  • Between-stage confirmation before continuing.
  • Defer all API ground-truth to marketing-studio; never restate or invent preset slugs, parameter shapes, or UUIDs here.
  • Keep internals silent unless the user asks how it works.

9. Source acknowledgment

Translated from Adil Aliyev's higgsfield-content-factory source skill (997 lines). The 5-stage pipeline, the UGC-first 5-format campaign taxonomy with concept seeds, the even-split allocation math, the button-driven UX discipline, and the per-batch generation gate are the net-new orchestration material the v3.7.13 marketing-studio translation deliberately deferred. Marketing Studio API specifics are not re-documented here — they live in ../higgsfield-marketing-studio/SKILL.md, which also reconciled several source-corpus API claims (preset slug naming, avatar handling, the absence of a generate_video.mode slug); this sub-skill honors those reconciliations and routes all API decisions through it. Default format shares and concept seeds are craft opinion, presented as overridable defaults, not platform rules.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Writes and refines prompts for Higgsfield AI's video and image models, and picks the model, camera controls and motion presets for each shot. Use whenever the user asks anything about Higgsfield AI — writing or refining video/image prompts, choosing a model (Kling, Veo, Wan, Seedance, Minimax Hailuo, DoP, Soul, Nano Banana, Seedream, Flux, GPT Image, etc.), camera controls, named motion presets, Soul ID character consistency, Cinema Studio 2.5/3.0, Vibe Motion, troubleshooting failed generations, credit optimization, Photodump, or any mention of higgsfield.ai. Also trigger on generic "write me a video prompt" or "make me an AI video prompt" requests when Higgsfield is the user's configured platform.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

Use when the user wants a 3D model, mesh or GLB out of Higgsfield — image-to-3D, multi-view-to-3D, text-to-3D, rigging or animating a mesh, remesh / retopology, retexture, or 3D Body (human shape + pose from one photo) — or mentions 3D Jutsu / Scene Builder 3D (private Blender scene projects built through Python). Also use for a 3D turnaround of a character or prop as a consistent multi-angle reference, or a physical 3D blockout rendered as a staging reference for a video shot. Names that trigger it: generate_3d, sam_3_3d, image_to_3d, multi_image_to_3d, meshy_v7_image_to_3d, tripo_h3_1_image_to_3d, tripo_h3_1_multiview_to_3d, hunyuan3d_v3_image_to_3d, meshy_v6_text_to_3d, hunyuan3d_v3_1_text_to_3d, tripo_3d, 3d_rigging, meshy_v5_remesh, meshy_v5_retexture, sam_3_3d_body, animation_actions, scene_builder_3d_*.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

Writes the character-performance layer of a video prompt as behavior under pressure, not displayed emotion — objective, obstacle, tactics, beats, subtext, listening, body/status/proxemics, and mandatory eye life. Produces a reusable 150–220-word acting master profile per character plus a per-scene rewrite of it, and a locked voice prompt. Use whenever a prompt needs acting, performance, or emotion direction; whenever characters read wooden, dead-eyed, or 'AI'; whenever a character must stay themselves across many shots; or when the user asks for character behavior, mannerisms, tics, a gait, or how someone reacts. Pairs with higgsfield-facs (muscle-level AU codes) and higgsfield-seedance (the prompt the paragraph goes into).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

Use when the user asks about Higgsfield's one-click Apps, wants to know which app to use for a specific output, or needs guidance on the Apps workflow.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

Use when the user asks about Higgsfield Assist (the built-in GPT-5 copilot), how to use the platform's native AI assistant, credit optimization strategies, plan selection, how to get more from fewer credits, or platform efficiency tips.

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

Use when the user asks about audio in Higgsfield videos, needs to add dialogue or lip-sync, wants sound effects or ambient sound in generated video, asks about music or BGM in output, or is using any audio-capable model (Kling 3.0, Seedance 1.5 Pro, Seedance 2.0, Veo 3/3.1, Grok Video / Grok Imagine). Also use when the user's prompt would benefit from audio direction but they haven't mentioned it. Also use when the user wants standalone audio — a soundtrack, ambience bed, multi-speaker scene audio (Seed Audio 1.0), or text-to-speech voiceover. Also use to swap or revoice the speaker in an existing video (voice_change), or to clone / create a reusable voice (create_voice → a voice_type 'element' voice usable in TTS and voice change).

한국어 요약은 준비 중입니다. 원문 설명을 표시하고 있습니다.

OSideMedia/higgsfield-ai-prompt-skill7802026년 10월 24일 업데이트

OSideMedia의 스킬 모두 보기

이 스킬의 문제 신고