awesome-claude-video-skills: A Safety-Rated Checklist for Picking Video Skills for Coding Agents

Before installing any video Skill, spending a few minutes reading this list first is the cheapest, lowest-risk first attempt — it doesn't run any code itself; it just helps you pick the right tool by video type and safety rating.

Who It's For

• Users who want Claude Code, Codex, or similar coding agents to create or edit videos but don't know which open-source skill, plugin, MCP server, or toolkit to choose.
• People who want to filter by use case: general frameworks and toolkits (29), promos and demos (26), explainer videos (32), editing (35), short-form and social media (19), digital avatars (4), story and animation (13), motion graphics (11), scripts and learning (9), music videos (2) — 168 repositories in total.
• People who care about safety prerequisites: every entry's README and installation steps have been safety-rated by Agent Skills Hub (e.g., SAFE, CAUTION); unrated entries show pending. A pending rating means the catalog hasn't issued a rating yet — read the repository yourself before using it.
• When it's not a fit: this is not a tool you can install and run; it cannot generate videos by itself, and it does not replace reading the specific repositories. Whether a given repository supports your agent, text language, or file formats depends on that repository's own documentation; this list does not cover each one individually.

Complete Your First Attempt in 3 Minutes

• Prerequisite: access to GitHub is all you need; no runtime installation required.
• Recommended primary path (read-only browsing): open the repository https://github.com/zhuyansen/awesome-claude-video-skills and read the category tables and safety-rating column in the README. This list has no package installer — this read-only entry point is the only recommended primary path.
• Low-risk task: in the README, find the category matching your video type, pick an entry rated SAFE, and hand that repository's README to your agent (e.g., Claude Code), asking it to read and report the installation steps, safety rating, and possible side effects. This step involves only reading and summarizing — nothing is installed or executed.
• Success signal: the agent can accurately state your selected entry's category, its safety rating (SAFE / CAUTION / pending), and the purpose declared in that repository's README. This step only proves information retrieval succeeded; it does not mean the chosen tool itself is safe to use — before installing, still verify against its official steps.
• Secondary note (for filtering and data purposes, not the primary entry): the official online filtering page https://agentskillshub.top/best/claude-video-skills/ refreshes every 8 hours; machine-readable data lives in data/skills.json in the repository (generation date 2026-09-27). To add an entry to the list, the official route is to open a GitHub issue and go through the same inclusion process as other entries.

Use Cases

Case 1: Compare options by safety rating. Scenario: you want to make a product explainer video (Explainers category, 32 entries). Give your agent this task: "Read the README of this awesome list and list the entries in the Explainers category rated SAFE, along with their purposes." The agent will answer based on the README's category tables and safety-rating column, giving you a candidate list with safety ratings; for pending entries, the agent can only tell you they're unrated — it cannot guarantee their safety for you.

Case 2: Batch-filter using machine-readable data. Scenario: you want to programmatically filter out all Avatars (digital avatars, 4 entries) repositories. Give your agent this task: "Read https://github.com/zhuyansen/awesome-claude-video-skills/blob/main/data/skills.json , filter by type, and summarize each entry's safety rating." Note that the file's generation date is 2026-09-27; entries added after that may not be included. Cross-check data freshness against the README or the page that refreshes every 8 hours. Both cases only read data and never modify your files.

Usage Boundaries

• Network: browsing the list requires access to GitHub; using the online filtering page requires access to agentskillshub.top; Case 2 requires the agent to be able to read repository files. When you have your agent download or visit any selected video tool repository, the scope of access depends on that repository and your agent's configuration; this list does not specify it.
• Side effects and confirmation points: browsing and filtering the list itself does not modify files or execute the list's code; but the video skills collected in the list generally generate videos and may execute rendering commands. Before installing, you must read the entry's README for its safety rating and installation steps, and confirm side effects before acting. What this list produces is rewritten recommendation information — it does not directly modify user files.
• Official limitations: inclusion criteria are the ability to create or edit videos, being agent-operated (skill, plugin, MCP server, or toolkit), and having a README; repositories with fewer than 50 stars must meet a higher README quality bar. The safety ratings rate each entry's README and installation steps — they are not a full verification of runtime code behavior.
• Unverified items: which agent versions a specific entry supports, which text or subtitle languages it supports, and whether it is cross-platform are all subject to that entry's own documentation; this list's evidence does not cover them. For installation differences across platforms, defer to each entry's official instructions; this list does not cover each one individually.