By cmgramse
Make media just by asking Claude in plain language — product images, social ads and A/B variants, short videos, a video of you talking, translating or dubbing a video, jingles, sound effects, reusable brand styles, face/voice identities, and animations / motion graphics (kinetic typography, lower-thirds, title cards, explainers, motion data-viz, 3D, Lottie). Examples: "make me a product image", "create a video of me explaining this", "translate this video into German", "make a jingle", "make an animated explainer". You don't pick a provider or run commands; Claude understands the request, grounds it in your brand and design, tells you what it will make and which credits it will spend, asks for a yes, then creates and saves the result. Under the hood it routes to the right provider (HeyGen for presenter/avatar videos and video translation; Higgsfield for images, product shots, short video, and audio; a local Remotion studio for deterministic, code-defined motion graphics), with lightweight pre-spend confirmation and no stored credentials.
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Use for any request to create, generate, or edit media — images, product photos, social ads, A/B ad variants, short videos, talking-head / avatar / presenter videos, video messages, translating or dubbing an existing video, music, jingles, sound effects, ambience, reusable visual styles, or face/voice identities. Use even when the user explicitly names HeyGen or Higgsfield. Just ask in plain language. This is the single entry point for making media: it understands the request, grounds it in the project's brand and design, makes the actual prompt or script, confirms before spending any credits, creates the asset, and saves the result. For web design systems use site-designer; for slide or carousel layouts use the template-mediated visual-asset bridge; for page or offer strategy use conversion-designer or funnel-strategist; for copy use content-creation-framework.
Internal provider executor for the creative-media-generation plugin — creates a persistent HeyGen avatar (a reusable face + voice identity) via HeyGen Avatar V, prompt-based by default with optional photo upload for real-person twins. Invoked by the orchestrator (creative-media-generation), not selected directly from user chat. The orchestrator handles request understanding, provider discovery, spend-safety, consent grounding, levers, and artifacts, then delegates the raw create/train here. If the user names HeyGen directly, route to the orchestrator first. Returns avatar_id + voice_id for the orchestrator to pass into heygen-video.
Internal provider executor for the creative-media-generation plugin — translates and dubs an existing video into another language with voice cloning and lip-sync (HeyGen Video Translation: same face, cloned voice, re-synced lips). Invoked by the orchestrator (creative-media-generation), not selected directly from user chat. The orchestrator handles request understanding, provider discovery, spend-safety, levers, and artifacts, then delegates the raw translate/dub here. If the user names HeyGen directly, route to the orchestrator first. Operates only on an existing source video (audio-only mode supported); returns the translated video/audio URL per language.
Internal provider executor for the creative-media-generation plugin — the HeyGen presenter / talking-head video renderer (v3 Video Agent: Frame Check, prompt engineering, avatar resolution, voice selection). Invoked by the orchestrator (creative-media-generation), not selected directly from user chat. The orchestrator handles request understanding, provider discovery, spend-safety, brand grounding, levers, and artifacts, then delegates the raw render here. If the user names HeyGen directly, route to the orchestrator first. Accepts avatar_id from heygen-avatar or a stock presenter; returns video share URL + HeyGen session URL.
Internal provider executor for the creative-media-generation plugin — the general Higgsfield renderer for images, video, 3D assets, audio (music/SFX), Marketing Studio ads, and Virality Predictor scoring. Invoked by the orchestrator (creative-media-generation), not selected directly from user chat. The orchestrator handles request understanding, provider discovery, model selection, spend-safety, brand grounding, levers, and artifacts, then delegates the raw render here. If the user names Higgsfield directly, route to the orchestrator first. Chains with higgsfield-soul-id for face/identity consistency under the orchestrator's direction.
External network access
Connects to servers outside your machine
This repository (skill-development) is the skill factory for iurFriend: the centralized environment where all Claude Code plugins are developed, tested, and refined. It is structured as a plugin marketplace (iurfriend-skills), but the primary activity is building and refining skills, not distributing them. Forward-going, marketplace-distributable skills ship as plugins (single- or multi-skill) per Anthropic's canonical shape.
First stop for anyone (AI or human) doing skill work — read these three before touching anything:
CLAUDE.md— the de-facto orientation doc: the full plugin catalog, plugin architecture, credential/contract/orchestrator doctrine, and the development workflow.PATTERNS.md— Layer-0 canonical index of every binding doctrine, principle, R-decision, and shared contract (routes to the sources in.agents/andconcepts/skills-ecosystem/).AGENTS.md— the Codex / Opencode parity entry-point; mirrors CLAUDE.md's catastrophic-invariant summary for non-Claude harnesses.The README below is the marketplace-facing front door. The canonical rules live in the three files above — this README orients, it does not govern.
25 marketplace plugins (plus the resident iurfriend-skill-patterns carrier, counted separately) across the marketing operating system and supporting web/management/observability tooling — brand extraction, funnel and conversion design, SEO, content creation, concept development, measurement and optimisation, site building/design/review/refactor, solution architecture, team and finance operations, data-collection engineering, creative media and podcast generation, voice, and LLM observability operators (LangFuse, MLflow).
Plus a vendored (third-party) tier — 18 vetted-external developer-tooling and productivity plugins imported for iurFriend employees (superpowers, beads, cloudflare, n8n-skills, supabase, obsidian-skills, defuddle, impeccable, extract-design, agent-browser, using-entire, frontmatter-generation, asana, and the official agent-sdk-dev / claude-code-setup / claude-md-management / feature-dev / mcp-server-dev). These are not factory-built: each is copied verbatim from a permissively-licensed upstream (MIT/Apache-2.0, or license flagged UNCONFIRMED), carries its LICENSE + UPSTREAM.md, and is listed in the auditable VENDORED.md register. The Anthropic docx/xlsx/pdf/pptx document skills are proprietary and not vendored; they are approved for install directly on iurFriend laptops — see EXTERNAL-PLUGINS.md (the companion register to VENDORED.md) for the exact install commands. Full vendored per-plugin catalog: the Vendored plugins section in CLAUDE.md.
For the authoritative, per-plugin catalog with one-line purposes, see the Plugins in This Marketplace table in CLAUDE.md. It is kept current; this README does not duplicate it.
Every top-level entry, one line each. Plugin directories are listed collectively (each is a self-contained plugin — see the canonical layout in CLAUDE.md § Plugin Architecture); the per-plugin catalog is in CLAUDE.md.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimnpx claudepluginhub cmgramse/skill-development --plugin creative-media-generationMLflow GenAI observability and evaluation operator — onboarding, tracing instrumentation, trace/session analysis, agent evaluation, metrics, and docs search.
Local podcast studio — turns any source (topic, URL, PDF, markdown, repo) into a finished branded episode: research, multi-voice script, Gemini TTS, music mix, show notes, and cover image. Ships the podcast-studio orchestrator skill and a /podcast command. Brand-as-config via show profiles; decoupled from any cloud control plane.
Rendering presenter — brainstorms, selects, and renders the most impactful visualisations for producer artifacts or ad-hoc data into self-contained (or interactive) reports. Emits frontmatter-free presentation siblings under presentation-sibling-contract@1. A strict presenter: it represents data and producer-supplied interpretations, never computes new findings, scores, deltas, or verdicts.
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Project bootstrap operating layer for Codex, Claude Code, and OpenCode: Beads setup, audit logging, AGENTS.md, CLAUDE.md, project rules, and orchestrator/executor/reviewer roles.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Complete creative writing suite with 10 specialized agents covering the full writing process: research gathering, character development, story architecture, world-building, dialogue coaching, editing/review, outlining, content strategy, believability auditing, and prose style/voice analysis. Includes genre-specific guides, templates, and quality checklists.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.