name	release-stats
description	Use this skill when the user wants a download/adoption report for PaceBar releases — totals per operating system, per release, or a raw CSV of GitHub release download counts.

Pull live GitHub release download stats and report adoption per OS. Backed by scripts/release-stats.mjs, which fetches data via the authenticated gh CLI (no token needed).

Run from the repo root:

bun run release:stats                        # report for cbnsndwch/pacebar
bun run release:stats -- --repo owner/name   # another repo
bun run release:stats -- --csv stats.csv     # also dump raw per-asset CSV

The report shows:

Fresh installs by OS — the real adoption number, with macOS arm64/x64 split. .dmg/.exe/.msi → installs; .AppImage/.deb/.rpm → Linux.
macOS fresh vs auto-update — .dmg (fresh) and .app.tar.gz (auto-update) are separate assets, so they split cleanly. On Windows the updater re-runs -setup.exe, so .exe = fresh + updates merged (not separable).
Latest release bundle sizes per asset.
Update activity — latest.json poll count = anonymous updater checks (every 15 min per running app), NOT installs and not version-tagged.

When reporting to the user, lead with fresh installs per OS. Always call out that latest.json is updater polling, not adoption.

Limits to remember: GitHub counts are cumulative, non-unique, and have no time dimension — for trends you must snapshot daily and diff.

Requires gh to be authenticated (gh auth status). If it isn't, tell the user to run gh auth login.

For ad-hoc breakdowns (per-version, date ranges, charts) beyond what the script prints, run with --csv and analyze the CSV directly.

name	release-stats
description	Use this skill when the user wants a download/adoption report for PaceBar releases — totals per operating system, per release, or a raw CSV of GitHub release download counts.