⏳ This skill is pending AI review.

Scores will appear once the review pipeline completes.

v0.6.0

cassette-model

cassette-editor@cassette-editor⭐ 220 stars· last commit 2d ago· 0 open issues

Show or change the Cassette editing model and thinking level for the current media session. Invoke only when the user explicitly asks for the model picker, wants to inspect the current Cassette model, or requests a model/thinking change.

Choose how to use this skill

You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.

1. Native installer

This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.

Do not guess an installer command or replace an existing version without reviewing the diff.

2. Complete package recommended

Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.

No complete ProSkills package is published for this listing yet.

3. Prompt-only

Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.

Need only the instruction file?

Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.

No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.

—/10

// RATINGS

⭐GitHub Stars
⭐⭐⭐ 220 on GitHubGitHub ↗

Popular

🟢ProSkills Score
—
📍

Not yet listed on ClawHub or SkillsMP

// README


One real Claude Code session, with every prompt shown as it was typed. Five 4K travel films (10½ minutes of footage) and a song go in, a two-sentence brief and a one-line revision go to the Cassette agent word for word, and a 30-second 1080p MP4 lands in cassette-exports/. Waits are fast-forwarded: in real time the import took 4½ minutes, the first cut 9½, the revision 3½ and the export 1.

Cut a 30-second teaser for Turkey from these five films, on the beat of Pamgaea.mp3. Open on a hot-air balloon lighting up at dawn in Cappadocia, travel from the ancient ruins to the sea, name each place when it first appears, and end on the misty lake with the title TURKEY.

The ruins part drags. Keep only the two best Ephesus shots, and put the camel right after the balloons.

Install in 30 seconds

You need uv (it fetches Python and the package on first start) and, to import video into a cloud project, ffmpeg.

1. Add Oh My Cassette to your agent.

# Claude Code
claude plugin marketplace add Cassette-Editor/oh-my-cassette
claude plugin install oh-my-cassette@cassette-editor
# Codex
codex plugin marketplace add https://github.com/Cassette-Editor/oh-my-cassette.git
codex plugin add oh-my-cassette@cassette-editor
# Hermes
hermes mcp add cassette --command uvx --args oh-my-cassette==0.6.0

For OpenCode or any other MCP host, see Install below.

2. Sign in, whichever host you use:

uvx oh-my-cassette==0.6.0 login

Your browser asks for the code Cassette emails you. Then restart your agent if it cannot refresh its tool list, and say:

Start a new Cassette project and import the clips in ./footage, then cut a 30-second travel video with a title at the start.

Projects and retention

A new project lives in the cloud unless you place it on one of your Macs; when your account has a Mac running Cassette for Mac, the service asks which you want.

  • A cloud project is deleted with its media after three idle days. Every use pushes that back, and each answer says when it is due (retention_expires_at). Exported files are already on your machine in cassette-exports/ and stay there.
  • A project held by Cassette for Mac is kept. The bridge can open, edit, import into and export it while that Mac is online. When the bridge runs on that Mac, the app reads each file where it is and nothing uploads; from any other machine each file uploads unchanged, with its embedded metadata, and the cloud copy is deleted once the Mac has it. docs/privacy.md has the details.

Overview

Oh My Cassette connects your coding agent to the Cassette editing agent. It is a local stdio MCP server that relays to the Cassette MCP service (CASSETTE_MCP_URL), which in turn calls the Cassette backend:

Claude Code / Codex / OpenCode / Hermes  ──stdio──▶  oh-my-cassette  ──HTTP──▶  Cassette MCP service  ──▶  Cassette backend
  • The tools are the service's. Whatever it lists (projects, editing turns, questions, undo, export) is what your agent sees, with the service's own descriptions and workflow instructions. When Cassette changes, the tools change with it; the plugin does not need a release.
  • The bridge adds only what needs your machine. Local files named in a call are sent first (only from the workspace, media types only). For a cloud project they are prepared here: video is transcoded with ffmpeg into the renditions the service declares, and audio or images the service cannot take as they are get converted. For a project a Mac holds they go as they are (see Projects and retention). Each file uploads while the next one is prepared, in parts, four transfers at a time; a file whose upload was interrupted picks up its earlier import when you import it again. Exported files are downloaded into cassette-exports/. The workspace identity travels with each call so the service can remember a project per directory.
  • Long work never times out the host. Every call waits a bounded time; slow local preparation answers preparing and keeps going in the background, and the next identical call picks it up.
  • It survives deploys. It reconnects on its own, retries a call that carries a request id when the connection drops or the service turns it away, resumes interrupted downloads, reports bridge.* errors with a retryable flag (and retry_after_seconds when the service said when), and, while the service is unreachable or refuses the token, lists a single cassette_bridge_status tool that says why.

The rules between the plugin and the service are defined in src/oh_my_cassette/contract.py; oh-my-cassette check verifies a service against them.

Requirements

  • An invited Cassette account. While the service is unreachable or refuses your sign-in, the only tool listed is cassette_bridge_status, which says why.
  • uv 0.5 or newer. uvx oh-my-cassette==0.6.0 resolves Python 3.11 or newer by itself.
  • ffmpeg (with ffprobe) to import video into a cloud project: brew install ffmpeg, sudo apt install ffmpeg or winget install ffmpeg. Video is transcoded on your machine before it uploads, and a build with zscale also tone-maps HDR footage properly. A project held by Cassette for Mac does not need it.

The service declares which file types each tool accepts and how they are prepared; the bridge sends only video, audio and images unless you set CASSETTE_UPLOAD_ANY_TYPE.

Configuration

Sign in once with oh-my-cassette login, using the Cassette account that holds your projects; the

// HOW IT'S BUILT

// FILES

2 files shipped with this skill. Inline text is the catalog copy. The upstream link is the source repository.

  • .agents/skills/cassette-model/SKILL.md1.5 KBupstream ↗
    ---
    name: cassette-model
    description: Show or change the Cassette editing model and thinking level for the current media session. Invoke only when the user explicitly asks for the model picker, wants to inspect the current Cassette model, or requests a model/thinking change.
    version: 1.0.0
    license: MIT
    metadata:
      tags: [cassette, model, thinking, mcp]
      category: media
    ---
    
    # Cassette model picker
    
    This is an explicit settings action. Never invoke it automatically during media ingestion or before
    an edit. A fresh Cassette session already uses GPT-5.6 Luna with `xhigh` thinking.
    
    1. Reuse the active Cassette `session_id` from the current conversation. If no Cassette media
       session exists yet, tell the user to add media first; do not invent an id or create an edit.
    2. Call `cassette_config` with only `session_id` to read the current selection and static options.
    3. If the invocation already names a valid model and/or thinking level, call `cassette_config` with
       those values and confirm the saved choice in one line.
    4. Otherwise show the model and thinking choices as two compact numbered lists and wait for the
       user's selections. Then call `cassette_config` with the selected model label and thinking value.
    5. Explain that the setting is scoped to this Cassette session and applies from the next edit turn.
    
    Selectable models are GPT-5.6 Luna and GPT-5.4 Mini. Thinking values are `off`, `minimal`, `low`,
    `medium`, `high`, and `xhigh` (display `xhigh` as “Extra High”).
    
  • README.mdupstream ↗

    Contents are not inlined in the catalog. Open the upstream file.

Plain-text copy, no login: /skill-bundles/cassette-editor-oh-my-cassette-agents-skills-cassette-model.md

// REPO STATS

220 stars
0 open issues
Last commit: 2d ago