asciinema-streaming-backup_skill

Overview

This skill streams asciinema terminal recordings to a GitHub orphan branch in real time, creating resumable zstd chunks and final brotli archives via GitHub Actions. It uses idle-detection to split recordings into sensible chunks and keeps recording history isolated from main repo history. The system is cross-platform for macOS and Linux and focuses on reliable, low-overhead backups.

How this skill works

The agent wraps asciinema recording with an idle-chunker that emits concatenatable zstd chunks into a local repo layout. Chunks are committed to an orphan branch and pushed to GitHub as they complete. A GitHub Actions workflow recompresses stored zstd chunks into a single brotli archive for maximal compression and long-term storage. Preflight checks, account detection, repo normalization, and configurable parameters guide setup.

When to use it

• You want real-time, incremental backups of asciinema recordings to GitHub.
• You need idle-detection to split long terminal sessions into manageable chunks.
• You prefer storage isolated on an orphan branch to avoid polluting main history.
• You want server-side recompression (brotli) via GitHub Actions for archival savings.
• You need a reproducible, scripted setup on macOS or Linux.

Best practices

• Run the preflight check to ensure asciinema, zstd, brotli, git, and gh are installed before setup.
• Use the 30s idle threshold default for a balance of chunk size and upload frequency.
• Store recordings in a dedicated repo or dedicated orphan branch to keep artifacts separate from code.
• Enable auto-push when network is stable; disable when working offline and push manually later.
• Keep zstd at moderate levels (level 3) for streaming-friendly compression; rely on Actions for final brotli recompression.

Example use cases

• Live developer sessions streamed and backed up so recordings are available immediately for postmortem or sharing.
• Automated CI job that records integration debug sessions and stores compressed archives for audits.
• Team productivity recordings captured on shared machines, split by idle periods to avoid huge files.
• Long-running demos where periodic chunk commits reduce risk of total loss on interruption.
• Academic or training labs that must preserve terminal sessions without mixing with source history.