agentspace-so/runcomfy-agent-skills
30 skills5M installs450 starsGitHub
Install
npx skills add https://github.com/agentspace-so/runcomfy-agent-skillsSkills in this repo
1Video EditVideo Edit is a RunComfy-focused agent skill that acts as a smart router between your edit intent and the correct model in the RunComfy catalog. Solo and indie builders who ship demos, ads, or social clips with Claude Code, Cursor, or Codex can describe transforms—general restyle, background or packaging swap with identity and motion preservation, precise motion transfer from a reference clip, or lightweight outfit restyle—and let the skill select Wan 2.7 Edit-Video, Kling 2.6 Pro Motion Control, or Lucy Edit Restyle accordingly. It emphasizes documented prompting patterns for each path and invokes the local RunComfy CLI rather than manual dashboard hopping. Use when you already have source footage and need dependable edits without burning credits on mismatched endpoints. It fits builders treating generative video as part of product storytelling, not as a one-off toy experiment.211kinstalls2Image To VideoImage-to-Video on RunComfy picks the right i2v model for each intent—HappyHorse for general animation, Wan 2.7 with audio for lip-sync, Seedance for multi-reference motion—plus documented prompting patterns. It avoids burning credits on the wrong endpoint. Solo creators use it when turning product shots or illustrations into short motion clips.210kinstalls3Image EditImage Edit is a RunComfy Pro Pack agent skill that acts as a smart router between your edit intent and the right model in the RunComfy catalog. Solo and indie builders use it when they already have one or many reference images and need dependable image-to-image results—swapping backgrounds, removing objects, inpainting masked regions, preserving identity across batches, or rewriting headline text inside the frame—without burning credits on the wrong endpoint. The skill maps triggers like image edit, i2i, or rewrite headline to Nano Banana Edit for batch identity-preserving work, OpenAI GPT Image 2 Edit for multilingual text and layout-heavy composition, Flux Kontext Pro for high-fidelity single-reference local edits, and Z-Image Turbo Inpaint for mask-driven precision. Execution goes through the local RunComfy CLI (`runcomfy run`), so it slots into Claude Code, Cursor, or Codex workflows as procedural knowledge rather than a separate app. Complexity is intermediate: you need the CLI configured and clear edit instructions, but you do not maintain four separate prompt playbooks yourself.210kinstalls4Flux KontextFlux Kontext Pro on RunComfy packages Black Forest Labs' precise local edit model with documented prompting patterns and routing guidance to sibling editors. Agents call the RunComfy CLI endpoint instead of naive prompts. Solo creators use it when they need high-fidelity single-image edits with tight prompt control.210kinstalls5Nano Banana 2Nano Banana 2 on RunComfy wraps Google's Gemini-family flash text-to-image model with prompting patterns for fast iteration, in-image typography, and predictable framing. It documents pricing tiers and when to escalate to Nano Banana Pro or GPT Image 2. Solo builders use it for quick generative comps without hand-picking RunComfy endpoints.210kinstalls6Nano Banana EditNano Banana Edit on RunComfy documents Google's image-to-image edit endpoint for identity-preserving changes, background swaps, and batch edits up to twenty inputs. It includes spatial prompting patterns and routing to sibling models. Solo creators use it when editing product shots or ad creatives at scale through the RunComfy CLI.210kinstalls7Happyhorse 1 0HappyHorse 1.0 on RunComfy packages the top-ranked text-to-video model with native 1080p and synchronized audio, plus schema guidance for duration and aspect ratio. Agents call runcomfy run happyhorse/happyhorse-1-0/text-to-video with model-specific prompts. Solo creators use it for high-quality short clips without manually comparing video model catalogs.209kinstalls8Wan 2 7Wan 2.7 on RunComfy wraps Wan-AI's flagship text-to-video model with prompting guidance for multi-reference conditioning, audio-driven lip-sync via audio_url, and smoother transitions. The RunComfy CLI handles endpoint calls with documented schemas. Solo builders use it when they need lip-sync or reference-guided video beyond generic t2v defaults.209kinstalls9Gpt Image EditGPT Image Edit on RunComfy packages OpenAI's ChatGPT Images 2.0 edit endpoint with prompting patterns for preservation, multilingual in-image text, and multi-reference layout work up to ten images. Agents call runcomfy run openai/gpt-image-2/edit via the CLI. Solo builders use it when headline rewrites or precise compositing matter more than fast batch edits.209kinstalls10Seedance V2Seedance 2.0 Pro on RunComfy documents ByteDance's multimodal cinematic generator—combining multiple image, video, and audio references with in-pass lip-sync for 4–15 second clips. Agents invoke runcomfy run bytedance/seedance-v2/pro with bundled prompts. Solo creators use it for rich short-form video when a single text prompt is not enough.209kinstalls11Flux 2 KleinFlux 2 Klein on RunComfy wraps Black Forest Labs' distilled fast Flux 2 variant with prompting guidance for sub-second generation, multi-reference brand styling, and step-count tuning between quick drafts and polished frames. Agents pick 9B or 4B endpoints through the RunComfy CLI. Solo creators use it when iteration speed beats maximum fidelity.209kinstalls12Kling 3 0RunComfy agent skill for Kling V3.0 video generation through documented CLI paths and tier/mode selection.186kinstalls13Codex PetEnd-to-end skill to hatch a Codex Pet from a reference image and drop it into ~/.codex/pets.174kinstalls14Ai Video GenerationRunComfy-centric agent skill that picks a video model for the goal and runs the minimal CLI command with per-model prompting guidance.154kinstalls15Ai Image GenerationRunComfy-powered AI image generation skill that selects models by intent and outputs ready-to-run CLI commands for create and edit flows.154kinstalls16Runcomfy CliCore RunComfy CLI skill teaching agents how to authenticate, discover models, submit jobs, and retrieve results from the command line.154kinstalls17Face SwapFace and character swap skill that picks the appropriate RunComfy video or image endpoint and documents how agents should invoke it.153kinstalls18Ai Avatar VideoRunComfy CLI skill for AI avatar and talking-head video: picks the right audio-driven model, ships prompting guidance, and one-line run commands for lip-sync, presenters, and voiceover-to-video use cases.153kinstalls19Video InpaintingAgent skill for video inpainting on RunComfy: remove objects, clean regions, and patch frames via CLI with automatic routing across Wan, Lucy Edit Restyle, and Seedream sequential edit paths.151kinstalls20Image InpaintingAgent skill to inpaint, remove, or replace masked regions in images via the RunComfy CLI.151kinstalls21Controlnet PoseAgent skill to condition image or video generation on pose, skeleton, depth, or motion references via RunComfy.151kinstalls22LipsyncAgent skill to lip-sync faces to audio via RunComfy, choosing avatar, resync, or generate-and-sync routes.151kinstalls23Video ExtendContinue or lengthen existing video clips using Veo 3-1 through the runcomfy CLI with documented prompting patterns.151kinstalls24Image OutpaintingExtend image canvas beyond the original frame on RunComfy, routing across identity-preserving edit models from a single agent skill.150kinstalls25Video OutpaintingExpand video canvas and change aspect ratio on RunComfy, preserving central motion via Wan edit routes or higher-quality ComfyUI outpaints.150kinstalls26RelightAgent skill that relights still images on RunComfy: direction, temperature, intensity, and mood via CLI-triggered endpoints.150kinstalls27Elevenlabs Music GenerationRunComfy CLI skill for ElevenLabs Music: structured lyrics and style prompts into full tracks or instrumentals for apps, ads, and games.150kinstalls28Ace StepACE Step on RunComfy is an agent skill that exposes StepFun-AI’s open-weights music foundation model through the `runcomfy` CLI. Solo and indie builders use it when they need royalty-friendly music without standing up their own GPU stack: describe the track with tags and optional structured lyrics, get up to four minutes of stereo audio, or repair and extend an existing file with inpaint and outpaint time ranges. The skill documents which endpoint to call, how prompting differs between base ACE Step and the 1.5 lyric model, and the cost profile that makes long background beds economical compared with premium vocal generators. It fits builders shipping prototypes, game loops, short-form video beds, or draft songs where iteration cost matters more than broadcast-master polish.142kinstalls29Ai MusicAI Music is a RunComfy agent skill that acts as a smart router over the platform’s music catalog instead of forcing you to memorize model IDs. When a solo builder asks for a song, instrumental, jingle, game loop, or a fix to an existing take, the skill interprets intent—premium sung hook versus budget bed, multilingual lyrics versus tag-driven instrumental, repair versus extend—and selects ElevenLabs AI Music Generation or the ACE Step family accordingly. It bundles each model’s documented prompting habits with the smallest viable `runcomfy run` invocation, including inpaint and outpaint when the user wants to salvage a chorus or grow a thirty-second idea into a two-minute cut. That makes it practical across prototyping sound design, marketing audio, and iterative content pipelines without bouncing between separate skill docs.142kinstalls30Gpt Image 2gpt-image-2 is a RunComfy Pro Pack agent skill that teaches coding agents when and how to call OpenAI GPT Image 2 (ChatGPT Images 2.0) hosted on RunComfy’s Model API. Solo builders use it to stop guessing prompt shapes and endpoint paths: the skill documents text-to-image and edit flows via `runcomfy run`, the model’s edge in legible in-image text and logo-like marks, multilingual typography, and strict instruction following, plus the three fixed canvas sizes and phrasing that preserves elements during edits. It also states explicit routing rules to sibling RunComfy skills when another diffusion model fits better for photorealism or style. Installation is oriented around the agentspace-so RunComfy skills package. Invoke when users name GPT Image 2 or want ChatGPT-class image generation without wiring a private OpenAI key—ideal for landing heroes, ad variants, and app store art generated from the same agent session that writes code.18.5kinstalls