Better Skill Recorder

BETTER SKILL RECORDER · PREVIEW V0.1.0

Creating Skills just got simplerShow the task. Talk it through.

Record your screen and explain what you’re doing. AI organizes the steps and generates a Skill—no need to write it from scratch. Review, export, and bring it to your agent.

Download for Mac ↓Install with an agent ↗

Apple Silicon (M series) · About 158 MB · No Copilot required

  • Local privacy protection
  • Automatic cleanup
  • Bring your own LLM
  • One-click Skill export
  • Fully open source
WORKFLOW / 01● CAPTURE

Make your know-how reusable

REAL WORK　→　CLEAR STEPS　→　SKILL

RECORD → REVIEW → EXPORT

You demonstrate. AI documents.

Start with work you know, not a blank prompt

  1. 01

    Record a task

    Press Option + Shift + R and work as usual. Optional narration helps the model understand your intent

  2. 02

    Review the result

    Analyze with your configured model, check the task and steps, and add your own judgment

  3. 03

    Export your Skill

    Choose a share-safe or personal package, then use it with an agent that supports the required capabilities

GET STARTED

Two ways to get started

Apple Silicon Macs are supported today; Intel Mac / Windows / Linux installers are not yet supported

Manual installation

Download, verify, move to Applications

  1. Download and verify

    Download the ZIP and SHA256SUMS.txt to the same folder. Run this command from that folder and confirm it reports OK

    shasum -a 256 -c SHA256SUMS.txtDownload checksums ↗

  2. Move to Applications

    Unzip and move the app into Applications. If macOS blocks this preview, follow the installation guide to allow it manually—do not disable system-wide protection

  3. Grant access and connect a model

    Allow screen recording and restart the app. Narration needs microphone access. Enter a model ID, BaseURL and API key in Model management, test the connection, then save

Full installation and usage guide ↗

Agent installation

Copy this prompt to an agent with terminal and local filesystem access

An agent can help download, verify and install. You still approve macOS security prompts and permissions, and enter your own API key. No security bypasses

CONTROL STAYS WITH YOU

Easy to start. Yours to configure.

Recording and redaction run locally. Analysis and generation use the model service you configure. You choose the LLM

Explore advanced settings ↗
Minimum recording
5 secondsShorter recordings are discarded with a notice
Audio size limit
100 MBOversized files prompt compression or splitting
Recording retention
30 daysExpired sessions go to Trash when idle
Keyboard shortcut
Option + Shift + RConfigurable, with conflict checks

A few things worth knowing

Do I need a subscription?

The app is free and open source. No GitHub Copilot subscription is required. Bring your own OpenAI-compatible model and pay its provider for usage. Tool calling is required; visual analysis also needs image input support

Where does my recording go?

Analysis and generation send relevant content to your configured model service. Safe recording attempts local redaction, but can miss or incorrectly mask information. Turning off visual analysis prevents image uploads, not analysis text. Review content before sharing

What are the current limitations?

This preview is not notarized by Apple and needs manual approval; updates may need new permissions. There is no auto-updater or system-audio recording. First-time narration may download a local transcription model. Scheduled Skills need a capable host agent; the app does not run scheduled jobs

How do I report a problem?

Include the app version, macOS version, chip and reproduction steps. Redact screenshots and logs first. Never upload API keys, personal configuration or private recordings

Report an issue ↗