Revid Agentic Skills

Drop-in agent skills that turn any link, blog, product, article, tweet, PDF, or
one-line idea into a finished short-form video — by calling the
Revid MCP server (which wraps
Revid Public API v3).

13 skills · works with Claude Code, OpenClaw, Codex, Cursor, Gemini CLI, and
any MCP-capable agent.

Install as a Claude plugin

This repo is a Claude plugin marketplace. The revid plugin adds the Revid MCP
server (30 tools, OAuth sign-in, no API key to paste) and all the skills below.

In Claude Code:

/plugin marketplace add tmakersas/revid-skills
/plugin install revid@revid

Then run /mcp, select revid and choose Authenticate to sign in with your Revid
account. In Claude (web and desktop), add the marketplace under Customize, Plugins,
Add marketplace, with this repo's URL, or connect only the server under Settings,
Connectors, Add custom connector, with https://www.revid.ai/api/mcp.

Try: "Find what is going viral about productivity apps this week, make a 30 second
video from the strongest hook and schedule it on my TikTok for Friday 6pm."

Install in Cursor or Gemini CLI

The same repo is a Cursor plugin (.cursor-plugin/plugin.json, mcp.json) and a Gemini CLI
extension (gemini-extension.json).

gemini extensions install https://github.com/tmakersas/revid-skills

Then run /mcp auth revid in Gemini CLI. In Cursor, install Revid from the plugin marketplace, or
add https://www.revid.ai/api/mcp under Settings, MCP. Both sign in with your Revid account.

• 🌐 Marketplace: https://www.revid.ai/skills
• 🧩 MCP server: https://www.revid.ai/mcp
• 📦 ClawHub: https://clawhub.ai/@api00
• 🛠 Source: this repo (MIT-licensed)

Install with the skills CLI

Works with Claude Code, Codex, Cursor, Gemini CLI, OpenCode and 40+ other agents
(skills.sh/tmakersas/revid-skills):

npx skills add tmakersas/revid-skills            # pick skills interactively
npx skills add tmakersas/revid-skills --skill revid   # just the main Revid skill

Then add the MCP server (step 2 below) so the skills have tools to call.

Install in 30 seconds

1. Get a Revid API key

Sign up at https://www.revid.ai/account and copy your key.

2. Add the Revid MCP server to your agent

Drop this once into your agent's MCP config (Claude Code's ~/.claude.json,
Cursor's mcp.json, OpenClaw's mcp.json, etc.):

{
  "mcpServers": {
    "revid": {
      "url": "https://www.revid.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_REVID_API_KEY"
      }
    }
  }
}

Any of these auth headers also works: x-revid-api-key, x-api-key, key.

That single MCP server exposes 30 tools, such as search_viral_videos,
render_video, get_project_status, export_video, calculate_credits,
schedule_publish and publish_now, used by every skill in this catalog. Claude,
ChatGPT, Claude Code, Cursor and VS Code can skip the API key: add the URL alone and
sign in with your Revid account (OAuth).

3. Install one skill — pick your agent

4. Ask the agent

Use Shopify Product Promo to turn https://allbirds.com/products/mens-tree-runners into a TikTok

The skill will scrape the URL, call render_video on the Revid MCP, poll
get_project_status until ready, and hand back an MP4 URL — typically in
2–4 minutes, ~40 credits.

No MCP support in your agent? Every skill still works against the raw
Revid Public API v3.
See the Direct HTTPS fallback section in
revid-api-foundations
and the examples/run.sh smoke tests.

The 13 skills

How a skill works

Every SKILL.md is plain Markdown with YAML frontmatter — no SDK, no runtime
lock-in. The shape is uniform across all 13:

---
name: revid-shopify-product-promo
description: One-paragraph "when to invoke this and what it does"
metadata: {"openclaw":{"requires":{"config":["REVID_API_KEY"]}}}
---

## When to use this skill
## Inputs
## Step-by-step                ← render_video → get_project_status → (export_video)
## `render_video` arguments    ← the JSON body to pass as MCP tool input
## Polling                     ← short note, full pseudocode in api-foundations
## Failure modes               ← table of error → fix
## See also

The agent reads the markdown, fills in {PRODUCT_URL} / {SCRIPT} /
{AVATAR_URL} / etc. from the user's request, calls the render_video MCP
tool, polls get_project_status, and returns the videoUrl.

Repository layout

.
├── skills/                  # the 13 SKILL.md skill folders
│   └── <slug>/
│       ├── SKILL.md         # the prompt agents load
│       └── examples/        # runnable JSON payload + run.sh
├── site/                    # Next.js marketplace UI (deployed to Vercel)
├── docs/                    # full Revid API v3 reference
├── examples/                # standalone curl / Python / TypeScript clients
├── shared/                  # cross-skill snippets (polling pseudocode etc.)
├── PUBLISHING.md            # how to publish updates to ClawHub
└── TODO.md                  # v1.1 backlog (watcher skills, agentic flows)

Local smoke test (no agent required)

Every skill ships a run.sh that exercises it via the direct HTTPS fallback
(plain curl against POST /api/public/v3/render), so you can verify a payload
without an MCP client at hand:

export REVID_API_KEY="rk_live_…"
cd skills/revid-shopify-product-promo
./examples/run.sh https://allbirds.com/products/mens-tree-runners
# → submits the render, polls until ready, prints videoUrl

To verify the MCP path itself, hit the JSON-RPC endpoint directly:

curl -sS https://www.revid.ai/api/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -H "Authorization: Bearer $REVID_API_KEY" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"calculate_credits",
                 "arguments":'"$(cat skills/revid-prompt-to-video/examples/honey-prompt.json)"'}}'

calculate_credits is free and confirms the payload shape — useful CI smoke
test without burning credits.

Publishing updates to ClawHub

Maintainer-only. See PUBLISHING.md for the full workflow.
TL;DR:

clawhub login
clawhub publish ./skills/<slug> --slug <slug> --name "Display Name" --version 1.0.1

Auto-updates on installed agents the next time they sync.

Roadmap

See TODO.md. Highlights:

• Watcher skills (revid-blog-watcher, revid-shopify-watcher,
  revid-news-watcher, revid-x-thread-watcher) — poll a feed, dedupe, fire
  the right base skill on each new item. Closes the "watch and react" gap
  identified in v1 review.
• Pro tier (revid-*-pro) using videoModel: veo3 / sora2 for premium
  output.
• Optional enrichment proxy that uses our OpenAI key to auto-craft
  scrapingPrompt + stylePrompt per URL — under evaluation.

Links

• Live site — https://www.revid.ai/skills
• Revid MCP server — https://www.revid.ai/mcp
• Revid API spec (direct HTTPS fallback) — https://documenter.getpostman.com/view/36975521/2sBXcGEfaB
• Revid Studio (build manually, then "Get API code") — https://www.revid.ai/create
• OpenClaw — https://openclaw.ai
• ClawHub — https://clawhub.ai

License

MIT. Fork the repo, adapt SKILL.md for any agent runtime — that's the point.