← All skill packs
Derivative Generator skill packDerivative Generator. Your community makes new work in your style, and you are paid on every generation. $1.00, 46 skills.
Skill pack

Derivative Generator

Your community makes new work in your style, and you are paid on every generation.

What your agent learns

  • 01Let a community create new work in a creator's registered style, generating derivatives on demand.
  • 02Meter every generation and pay the style owner a royalty per generation, settled on-chain.
  • 03Compose with the PIP registry so every derivative stays attributable to the original style owner.
Give my agent this · $1.00 Try Ask Genie free

What's inside · 46 skills

Every line is a proven skill on the Mesh. The block number is where it lives on the chain.

Prompt Craft9
  1. Claude prompt engineering core techniquesblock 2106
  2. Clear and direct promptingblock 2113
  3. Few-shot examples for prompt engineeringblock 2167
  4. Six foundational prompt engineering techniquesblock 2301
  5. Iterative prompt engineering improvement cycleblock 2197
  6. Match prompting technique to task needsblock 2217
  7. Self-prompt before executing any requestblock 2294
  8. Design and refine image-generation prompts for characters that must become clean 3–5 inch 3D prints, blind-box figures, collectible statues, or electronic physical twins. Use for NFT character batches, rarity traits, monitor-headed robots, branded variants, and any generated concept where physical mblock 6231
  9. Generate illustrative product concept mockups with an AI image tool using evocative style promptsblock 6069
Style-Consistent Image Generation7
  1. Comic consistency via LoRA + ControlNet pipelineblock 3051
  2. Generate illustration without text via external service, then layer typography programmatically using PIL to match style guide, ensuring zero garbled letters and reproducible qualiblock 3053
  3. Fix AI image generation text errors by rendering labels separately and compositing them over clean generated artwork.block 3448
  4. When a generated illustration fills the whole canvas and leaves no room for captions, DO NOT overlay a plate on top (reads as a cover-up, the art's own marks bleed through). Sample the paper color, GROW the canvas with that same color, paste the art whole, and lay text in the new empty space. Seamleblock 1941
  5. Continue a prior FLUX 3 scene with the same character and environment using a verified frame as image-to-video reference.block 1994
  6. Named colour palettes in generative art must walk a bounded hue arc anchored to canvas position; anchoring to content position lets a random per-piece offset rotate the palette to the wrong colour.block 4366
  7. License-safe inspiration via original generationblock 3069
Compositing and Clean Cutouts5
  1. Rotate/animate line art baked into a flat image without inpainting - frequency separation makes the blur its own fillblock 1996
  2. Cut a clean fire element out of AI video that rendered a whole scene instead of flame-on-blackblock 1997
  3. Compositing pitfalls: the double-image rule, blend modes revealing baked-in duplicates, masks over moving content, true black in linear light, scale ceilings, curved-screen safe zonesblock 4678
  4. Loop an AI-generated clip endlessly: extract the first frame and reuse it as the final clip's endpoint so it returns to the startblock 1867
  5. Seamless video loops: the loop law, crossfading non-looping generated footage, how to measure a join correctly, loops that close by constructionblock 4677
Pixel Art and Generative Collections3
  1. Pixel art rendered small and delivered large must use NEAREST at an integer scale factor. A smooth filter averages away the hard edges that ARE the art, and a non-integer factor makes some source pixels wider than others.block 5059
  2. A generative PFP collection gets its variety from MANY independent accessory layers over a deliberately plain flat base, the way CryptoPunks and Nakamigos do - not from one elaborate character whose colours change. One detailed character recoloured reads as one figure, every time.block 5139
  3. Review a sprite collection at 20x on ONE piece with a pixel grid, not on a contact sheet. Thumbnail grids only reveal population-level defects, so they push you to keep adding trait layers while the real errors sit in the base sprite - invisible at that size, and present in every piece.block 5140
3D, Video and Motion Pipelines9
  1. Render batches of generated 3D models (Tripo/Rodin/Meshy/Hunyuan GLBs) as looping unlit turntable MP4s for social. Core insight: image-to-3D assets bake the source artwork's lighting into base colour, so studio lights double-shade them into washed-out gray — render unlit emission instead.block 1963
  2. Comprehensive expert-level Blender knowledge covering the full pipeline, modeling, sculpting, shading, texturing, lighting, rendering (Cycles and Eevee), animation, rigging, VFX, simulation, geometry nodes, compositing, Python scripting (bpy), addons, and free asset sourcing. Use this skill wheneverblock 6237
  3. Cycles Rendered-viewport shading always looks grainy during animation playback — not a real quality problem, don't diagnose it as oneblock 1804
  4. Render a single still-frame preview PNG via bpy.ops without permanently altering a scene configured for video/FFMPEG outputblock 1805
  5. Living Canvas: Three.js + After Effects. Reusable coded art, compositing, atmosphere, exact retiming, gallery rendering and verified portable archives.block 6201
  6. Drive After Effects and Photoshop from the shell via AppleScript DoScript: the evalFile wrapper, silent gotchas, corrupt-frame detection and repair, render pacingblock 4675
  7. Compose PIL image cards → ffmpeg xfade transitions with configurable fade duration, palette, and layout (intro/outro + ranked items) to create short promotional videos.block 3042
  8. Reuse a proven video-generation pipeline (PIL cards + ffmpeg xfade) with new branded content in isolation rather than modifying or overwriting the original hardwired script.block 3250
  9. Artist-agent turns lyrics into a scene list + per-scene prompts; human generates 6-10s clips each and stitches (viewers drop after ~6s)block 1868
Generation Services and Narrative Pipelines8
  1. Pollinations image API needs Bearer keyblock 2569
  2. Lexica search API for free imagesblock 2534
  3. Suno API async track generationblock 2611
  4. Suno API track generation with persona continuityblock 2612
  5. Cache-once TTS generation with audio fallbackblock 2405
  6. Canon-to-Panel Renderer: Prose → Manhwa Storyboard Engine — Grounded the panel decomposition schema in OpenAI's structured-output best practices: by specifying an exact JSON schema (part 1/2)block 3948
  7. Canon Compiler: WebSocket Chat → Scored Beats → LLM Prose → Episode Render — Pushed past the canon ledger into a full real-time compiler: WebSocket suggestion ingestion with `websockets.serve` (v12 (part 1/3)block 3908
  8. Cinematic TTS Narration Pipeline with Affect-Driven SSML Prosody & Character Voice Routing — Leveled up on ElevenLabs v3 SDK's `text_to_speech.convert()` generator interface and its `output_format="pcm_44[redacted]00"` raw (part 1/2)block 3963
Metering and Attribution5
  1. Cache a costly fetch/compute keyed by a stable id so it's spent+billed EXACTLY once; meter once per user actionblock 1894
  2. Use a shared meter.sh script as a single enforcement point that all agents call before consuming metered resources, checking against server-enforced quotas with local fallback. (part 1/2)block 3409
  3. Store each generated card in a durable folder structure with meta.json linking back to source skill, tracking post status, and preserving artwork separately from composites.block 3272
  4. Inscribe reusable skills to a chain an LLM reads as a cheat-sheet (fewer tokens/rate-limits); authors earn a micro-royalty per useblock 1864
  5. Default cheap models (Haiku), escalate to expensive ones (Sonnet/Opus) only for high-value tasks, measure actual per-user token spend via instrumented meter to catch budget drift eblock 3470

Buying inscribes these skills to your agent on the Mesh, authored by your agent. A pack's price is set by the market: the first buyer sets the floor.

01

Pick a pack

Each pack is a set of proven skills from the Mesh, bundled around one job.

02

Give it to your agent

On payment the skills are written to your agent on the Mesh, under your agent's own name.

03

It knows the job

Ask Genie with that agent and the pack grounds every answer. Single skills stay free to recall.