pexoai/pexo-skills
5 skills53.5k installs3.7k starsGitHub
Install
npx skills add https://github.com/pexoai/pexo-skillsSkills in this repo
1Pexo AgentPexo Agent is an integration skill that connects your coding agent to Pexo’s AI video creation service. Solo builders use it when they need product ads, brand stories, explainers, or social clips without learning a timeline editor. You send the user’s request through the documented API flow; Pexo may ask clarifying questions or show preview options before rendering. Outputs are short videos between five and one hundred twenty seconds in common aspect ratios. Setup centers on a local config file with API credentials and lightweight shell tooling. It fits indie founders who want launch-week creative on demand while staying inside Claude Code, Cursor, or Codex. Review the Security Audits panel before handing production API keys to an automated agent.33.3kinstalls2Videoagent Video Studiovideoagent-video-studio packages an agent-facing path to short-form AI video from text or reference images. The skill description positions it as zero local API keys by routing generation through a hosted proxy that configures the fal.ai client on the server. Request bodies accept mode, prompt, optional imageUrl, duration, aspectRatio, and a model key resolved against a shared models registry. Supported backends span multiple vendor slugs such as minimax, kling, veo, hunyuan, pixverse, grok, and seedance, which helps solo builders compare outputs without rewriting integration code for each provider. Operators still supply FAL_KEY and VALID_TOKENS in the proxy environment and should treat rate limits and IP quotas as production concerns. Use it when your agent workflow needs programmatic video drafts for marketing clips, demos, or social content rather than manual dashboard exports.10.1kinstalls3Videoagent Image StudioVideoagent Image Studio is a generative-media skill for solo builders running video or creative agents that need reliable still frames from multiple commercial models without hand-rolling poll loops. Version 2.0 streamlined the operator story: invoke the packaged script and wait for a finished result while polling stays inside the tool. Outputs normalize to one JSON-friendly structure so downstream agent steps can branch on success, model id, primary imageUrl, image arrays, and jobId regardless of whether fal.ai or Legnext served the job. You can use the default hosted proxy or point IMAGE_STUDIO_PROXY_URL at your own deployment, with server-side keys for Flux-class models and Midjourney respectively. It suits agent pipelines in Build where thumbnails, storyboards, or key art must be produced repeatably as part of a larger video workflow.5.4kinstalls4Videoagent Audio StudioVideoAgent Audio Studio is an agent skill that centralizes speech, music, sound effects, and voice cloning for solo builders and small teams producing video or multimodal content. Instead of juggling separate ElevenLabs and fal.ai integrations, you describe what you need—narration, low-latency TTS, background music, SFX, or a cloned voice—and the skill routes the job to the appropriate model and returns a hosted audio URL. It targets creators using video agents or automated pipelines who already have an ElevenLabs API key and optionally install the @elevenlabs/mcp server. Use it during build when you are wiring media steps into an agent or content workflow, not as a substitute for storyboarding or legal clearance on cloned voices. Complexity is intermediate because environment setup and API billing apply.3.7kinstalls5Seedance 2.0 PrompterSeedance 2.0 Prompter is an agent skill for solo builders and small teams who generate video through Seedance 2.0 instead of guessing prompts in chat. Install it when you already have multiple reference assets—still frames, motion references, or audio—and need cinematic or shot-level direction translated into model-ready language. The skill acts as an expert prompt engineer: it ingests creative intent and file roles, then emits a structured JSON payload with the final prompt and parameter hints for whatever skill or tool invokes the model next. That keeps multimodal references aligned with motion, framing, and mood rather than dumping filenames into a generic template. Use it for ad spots, social clips, storyboards-to-video experiments, or any workflow where quality depends on deliberate prompt construction. It does not render video itself; it is the specification step immediately before generation in a chained agent workflow.951installs