Cumora 0.2.1
Two fixes to things that looked like features but never worked.
Desktop sign-in
Signing in from the desktop app handed the session back over the cumora:// URL scheme and left the rest to the OS. That is fine on a clean machine and unreliable everywhere else: any other build that ever registered the scheme — an old release/ build, a mounted DMG, a stray npx electron — can win it and swallow the token, so the browser says "signed in" and the app never hears about it.
The sign-in page now hands the token straight back to the process that started the sign-in, over the loopback connection it was already served from. The URL scheme stays as a fallback for the one case that path can't cover (the app quit while the browser was still on the provider). The single-use nonce and same-origin checks are unchanged.
A sound for every Skype emoticon
The emoticon sounds were wired up end to end — viewport trigger, click to replay, the preference toggle — against audio that was never added, so the whole feature was silent from day one. Nine emoticons now use the genuine Skype event sounds ((skype) the message pop, (call) the ring, (hi) the login chime, and friends).
Classic Skype never had per-emoticon audio, so the remaining 98 are synthesized: each emoticon gets a semantic family — sparkle for the cheerful ones, a falling minor for the downcast, glass for the drinks, a fanfare for the heroes, a whoosh for the ninjas — and its own pitch within that family, so (party) and (clapping) are recognisably related without being the same sound. scripts/gen-skype-sounds.py generates the set deterministically from the catalog and ships in the repo, so it can be re-run, re-tuned, or replaced.
Turn them off under Preferences → Skype emoticons.
Desktop app auto-updates. No server or CLI changes in this release; cumora on npm stays at 0.2.0.