gooseworks-ai/goose-skills

Deterministic video assembly pipeline for 3D character explainer ads. Trims clips, normalizes to 1080x1920, concatenates, and muxes audio with static-still fallback for missing clips.

Ad Angle Miner extracts ad angles from reviews, Reddit, support tickets, and competitor ads. Use it when you need a ranked bank of pain points, outcome phrases, competitor weaknesses, and proof quotes for campaigns.

Ad Campaign Analyzer turns Google, Meta, and LinkedIn performance data into campaign diagnostics and budget shift recommendations. It identifies winners, waste, statistical significance, and where to cut, scale, or test next.

Ad Lead Quality Analyzer joins ad platform data with downstream funnel events to replace signup CPA with true CAC per qualified lead. Use it to find tracking gaps and classify creatives as Scale, Keep, Investigate, or Cut.

Ad-To-Landing Page Auditor checks whether ad copy and landing pages carry the same promise, language, and CTA through the click path. Use it to find message mismatches that hurt conversions on Google, Meta, and LinkedIn campaigns.

Deterministically renders a viral iOS AirDrop notification-carousel video ad from a brand line and product photos, using headless Chrome, PIL, and FFmpeg. Ideal for creating crisp, real-product carousel ads without generative AI smear.

Deterministically assembles a cosmic-mythology voiceover reel from a config: weighted Ken Burns zooms over cosmic stills, synced to a spoken VO, with a hook line overlay and burned captions, using FFmpeg. Use to produce faceless cinematic storytelling videos with a myth-as-teacher reframe.

Generates or edits images using FAL AI models (e.g., nano-banana, gpt-image, flux) with calls routed through a proxy for billing. Ideal for creating keyframes, product hero images, and flat-cover transforms.

Generates an instrumental music bed using ElevenLabs Music, routed through a proxy for billing. Trims sparse intros, applies loudness normalization, and fades tails. Designed for music layers in video ad formats.

Assembles a picture-in-picture product-listicle video ad from a config, keeping the creator full-frame with overlays including a demo PiP, title pill, and product card. Builds per-beat overlay PNGs, composites onto creator clips, concatenates beats, and burns captions deterministically.

Generates voiceover clips using ElevenLabs text-to-speech, routed through a proxy to ensure proper billing. Used for spoken narration in video ad formats such as app sizzles and explainers.

Generate a single photoreal or designed image with OpenAI GPT-Image models via FAL.ai, supporting text-to-image and editing with reference images for character work, scene keyframes, and storyboards.

Generates video from image or text using FAL AI models (Kling, Seedance, Veo). Routes through GooseWorks proxy to bill the Ads agent; used for creating base clips in video ads.

Detects recent funding events across companies, qualifies them against your context, finds relevant contacts, and drafts personalized outreach. Tool-agnostic and works with any company source, contact finder, or outreach platform.

Scrapes competitor ads from Google Ads by domain, returning ad creatives, formats, and campaign details for competitive research and messaging analysis.

Routes all paid media generation (FAL image/video, ElevenLabs music) through the GooseWorks proxy to ensure billing goes to the Ads agent. Includes crash-resume to prevent double-billing.

Scrape competitor ads from Meta's Ad Library (Facebook, Instagram, Messenger, Threads, WhatsApp) by company name or page URL. Returns ad creatives, spend estimates, reach, and campaign details for competitive analysis.

Model Comparison Grid renders a dark, deterministic comparison video from a config with a centered prompt, staggered 2-4 panel grid, and minimal end card. It is used for same-prompt, multi-model showcases with image or video cells and frame-stepped playback.

Narrated UGC Wardrobe Stitch assembles a deterministic stitch-reply ad from a locked VO, EDL windows, clips, captions, music, a landing-page scroll, and a real end card. Use it to render the final vertical master from already-generated assets.

Discovers newsletters for sponsorship opportunities based on target audience and industry keywords. Uses web search, directories, and competitor intelligence to find and rank newsletters with estimated audience and sponsorship rates.

Assemble a premium 3D product-showcase ad from a config using four beat clips, deterministic Playwright end card, and an instrumental music bed. Use to create free, repeatable video ads without AI-rendered brand text.

Assembles absurdist animated explainer video ads from provided clips, voiceover, and music using Python, ffmpeg, and PIL. Handles retiming, re-encoding, concat, audio mixing, and captions burning for a deterministic 9:16 vertical video.

Produces a brand identity reveal video from a config, compositing HTML mockups into a real-space poster frame with music. Automates the hard-cut sequence and shadow integration for a behind-glass effect.

Assembles a 3D-CGI app sizzle video: AI-generated phone plates with real UI compositing, burst overlays, Ken-Burns fallback, and audio mix. Use to produce an Apple-keynote-style product reel where on-screen UI is always real assets.

Assembles a cartoon music video ad from a configuration that specifies the song, per-bar video clips, captions, and logo. Cuts each clip to its bar window, hard-concatenates, burns VEED-style captions at the bottom, overlays a persistent logo bug, appends a solid-color end card, and muxes the song underneath — all deterministically and for free.

Assembles a ChatGPT chat-reveal video ad from a thread and timeline JSON using Playwright and ffmpeg. Use it to record the chat, stream the assistant answer, and stitch in an end card and optional music bed.

Assembles a cinematic music-video ad by cutting i2v clips to lyric windows, hard-concatenating on the beat, muxing the anthem, burning lower-third captions from song word timings, and overlaying a brand end card — all deterministically with FFmpeg, no paid calls.

Assembles an editorial-motion podcast-clip ad using deterministic ffmpeg ken-burns, hard cuts, audio mux, captions, and a PIL end card. This free stage animates static keyframes on spoken beats without generative video.

Assembles a flat-vector product explainer video by compositing AI-generated character clips with real product photos using Remotion, PIL, and FFmpeg. Cut down from the animated master, never a static intermediate.

Assembles a wordless macro-tabletop food-product sizzle ad from pre-generated keyframes, videos, and music, applying anti-AI grain, compositing audio and a static end card using PIL. Use for fast, deterministic assembly of photorealistic food ads.

Deterministically assembles a multi-scene GRWM beauty demo ad from pre-generated assets, including voiceover-driven scene cuts, product overlay cards, captions, and music mixing, producing a ready-to-publish video at zero cost.

Assembles a product-hypermotion video ad with kinetic typography: dices one Seedance clip into segments, intercuts with PIL-rendered spec cards, and caps with a real-logo end card. All free steps, deterministic, for high-production sizzles.

Render IG Live Gallery turns a config and real product stills into an Instagram Live-style gallery video with branded chrome, comment feed, floating hearts, and a logo endcard. Use it for deterministic product-story videos with silent frames that can be muxed with music afterward.

Render IMessage Cascade assembles a 9:16 iMessage notification-cascade video ad from a phone plate, 3–5 messages, and an end card. It uses PIL and ffmpeg to draw authentic Apple Messages banners and animate the stack upward, then clears it and fades to the end card.

Render IMessage Chat assembles an iMessage chat-reveal ad from thread JSON with Playwright recording, ffmpeg stitching, SFX, and an optional music bed. Use it when you need a typed conversation animation that ends on a designed brand card.

Render Mosaic Grid Reveal builds a full-bleed mosaic-reveal video from config using real product stills, a real wordmark, and approved copy. It uses Playwright and FFmpeg to frame-step the scene for crisp text and deterministic output.

Render Multiworld assembles a silent, music-led 9:16 product-tour ad from trimmed world clips, an HTML end card, and a single music bed. Use it when you need deterministic FFmpeg/Playwright assembly for the multiworld-product-tour format.

Render Myth Vs Fact assembles a deterministic myth-vs-fact kinetic-typography explainer video from scripted copy, palette, VO, captions, and a brand end card. Use it for sound-off-safe explainer ads with beat-snapped timing and no AI-generated visuals.

Render Offer Ad creates a deterministic 12-second vertical offer ad from config, a real product photo, and typed-on-screen copy. Use it for motion-graphics offer ads that need beat-synced cuts, claim checks, and a 1:1 crop output.

Render Photo Grid Card builds a deterministic promo-card video from config with a fixed header and footer, a continuously scrolling 2-row mixed-media grid, and real DOM text. Use it for the photo-grid-promo-card format when you need frame-stepped assembly with FFmpeg and Playwright.

Render Podcast Skit assembles a two-host fake-podcast ad from per-line lipsync clips, white captions from VO timestamps, and a brand end card. Use it for the free deterministic assembly stage of a podcast-skit format.

Render Proof Points Overlay builds deterministic PIL/FFmpeg pill overlays for a 9:16 UGC ad format and composites them onto a base clip with a music bed. Use it when you need a config-driven proof-points cascade with crisp text and no paid generation calls.

Render Search Grid builds a deterministic 1080×1920 video from a config using real product and brand assets. Use it for the search-grid format with staged motion beats, captions, and optional music muxing.

Render Song MV assembles a sung-track music video ad from a config, cutting one visual tableau per lyric beat and syncing captions to the song's own word timings. Use it for the song-driven-music-video format when you want deterministic final assembly, including the end card and FFmpeg muxing.

Render Split Screen Creator assembles a split-screen creator ad from a config, combining a top product/demo clip with a bottom creator lip-sync take and a divider. It then transcribes the assembled cut and burns word-level captions for the final master.

Render Value Prop creates a config-driven value-prop video with a hook, short benefit claims, product visuals, and an end card. Use it for sound-off product spots rendered deterministically from Playwright and FFmpeg.

Render Vignette assembles short-form product vignette ads from cutouts, kinetic background video, text cards, and a separate music mux. Use it when you need a silent-first, loopable ad format with layered overlays and standardized end cards.

Stop-Motion Hand Swatch Cycle assembles a config-driven product-demo video from still plates, hard cuts, an HTML-rendered end card, and music. Use it to build or re-cut the stop-motion hand-swatch-cycle format without regenerating the source plates.

UGC Fix Loop stitches a replacement clip into a single-take UGC master with FFmpeg and vets a Seedance prompt through an OpenAI proxy. Use it to repair one beat without changing the master audio or to review a prompt before render.

VO Anchored Motion Listicle assembles a deterministic motion-graphic listicle video from a config, using word-level voiceover timing to anchor web-rendered beats, captions, B-roll windows, and final audio/video compositing. Use it when building a spoken listicle ad with Playwright-rendered scenes.

Verify Product Image checks a generated static ad image opens, matches the requested dimensions, and shows the correct product or subject. It also flags garbled text, wrong branding, and severe artifacts, then records the result in verification.md.