automate-this

Record yourself performing the task, then provide the video file path. The Skill extracts frames with ffmpeg, transcribes narration with Whisper, reconstructs your workflow step-by-step, and proposes automation scripts at multiple complexity tiers.

ffmpeg is required for extracting frames and audio from the video file. Whisper (openai-whisper or whisper-cpp) is optional and only needed if the recording contains audio narration you want transcribed.

Yes, visual-only analysis works by examining extracted frames to identify applications, actions, and data flow. Narration adds context about pain points and intent, but the Skill proceeds with frame analysis alone if Whisper is unavailable.

The Skill first checks for a public API since API calls are more reliable than UI automation. It falls back to curl for simple HTTP requests, then Playwright or Selenium for workflows requiring actual UI interaction.

Confirmation prevents building automation on a wrong understanding of your process. The Skill presents a numbered step list of what it observed and waits for your approval before proposing any automation code.

Credentials are never hardcoded in generated scripts. The Skill uses environment variables, macOS keychain access via the security command, or runtime prompts to handle authentication and secrets safely.