Tcut Terminal Video

Create a concise terminal story that runs reliably, proves the important output, and produces the requested media without hiding failures.

Core contract

Treat every command in a tcut script as a real shell command with the caller's permissions. t.hide() removes footage; it does not sandbox, redact logs, or make a command safe.
Prefer a scripted .video.ts file for agent work. Use tcut rec only when live, human-driven interaction is part of the request or when converting an exploratory session into an editable script.
Wait for observable terminal state before acting. Use t.run(), t.wait(), and t.expect(); use t.sleep() for deliberate pacing, not synchronization.
Assert the key result on the rendered terminal screen. A video that looks plausible but shows the wrong state is a failed demo.
Keep the story short. Show only the commands and outputs needed to explain one workflow, and use t.title() or t.print() only when a caption materially improves comprehension.
Preserve the target repository and its instructions. Do not install packages, start external services, modify product files, upload media, or force a re-record unless the user has authorized that action.
Never place tokens, credentials, private paths, or sensitive output in the script, cast, rendered media, shell history, captions, or publishing configuration.
Do not claim cross-platform or visual success without checking the produced artifact. tcut's Linux and Windows binaries may behave differently from its verified macOS path.