tv-watcher - SKILL.md Agent Skill

name: TV Watcher description: Track TV series and movies, get notifications about new episodes, date changes, and streaming availability. user-invocable: true

TV Watcher

Track TV series and movies. Get notified about new episodes, date changes, and streaming availability.

Tools

Tool	What it does
`watch_search`	Search for shows/movies by name
`watch_add`	Add to watchlist
`watch_remove`	Remove from watchlist
`watch_list`	Show current watchlist
`watch_status`	Show status and upcoming episodes for a title
`watch_recommend`	Get recommendations (requires TMDB or Trakt)
`watch_check`	Run checker: poll sources, find new episodes/releases, return report
`watch_set_locale`	Set interface language
`source_list`	List data sources and their status
`source_enable`	Enable a data source
`source_disable`	Disable a data source
`source_set_key`	Set API key for a source (DM only!)
`platform_list`	List user's streaming platforms
`platform_add`	Add a streaming platform
`platform_remove`	Remove a streaming platform

Natural language understanding

Understand user intent and call the right tool. Do not require exact syntax.

User says	Tool
"add Fallout", "track Shogun", "add to watchlist"	`watch_add`
"remove Fallout", "stop tracking"	`watch_remove`
"what in watchlist", "my list", "what am I watching"	`watch_list`
"status Fallout", "what episodes", "what is new"	`watch_status`
"recommend something", "what to watch"	`watch_recommend`
"find Fallout", "search series"	`watch_search`
"enable TMDB", "add source"	`source_enable`
"add Netflix", "platform add disney-plus"	`platform_add`
"language uk", "switch to Russian"	`watch_set_locale`
"check for updates", "any new episodes?"	`watch_check`

Context resolution

"add the second one" after search/recommendations - use the second item from the previous response.
"remove this" / "delete that" - resolve from conversation context.
"status" without a title - if only one item in watchlist, use it; otherwise ask.

Setup / Onboarding

When the user first asks about tracking series or activating TV Watcher:

Sources: TVMaze works without an API key (series only). Ask if the user wants TMDB, OMDb, or Trakt for movies and recommendations. These need free API keys.
- If yes: ask for the key, call source_set_key, then source_enable.
- Remind: API keys should only be shared in DM, not in group chats.

Notifications: Ask if the user wants automatic notifications about new episodes.

If yes: ask how often (every morning, twice a day, etc.) and where to send (current chat or another).

Create a cron job using the built-in cron.add tool:

{
  "name": "TV Watcher check",
  "schedule": { "kind": "cron", "expr": "<user's schedule>", "tz": "<user's timezone>" },
  "sessionTarget": "isolated",
  "payload": { "kind": "agentTurn", "message": "Call the watch_check tool and return its result verbatim. Do not add commentary." },
  "delivery": { "mode": "announce", "channel": "<channel>", "to": "<chat_id>" }
}

If the user says "send here", use the current channel and chat from the conversation context.

Platforms: Ask which streaming platforms the user subscribes to (Netflix, Disney+, etc.) and call platform_add for each.
Language: Detect from the user's language in chat. If they write in Ukrainian, call watch_set_locale with uk. If Russian - ru. Default is en.

Changing notification settings

"change notifications to 7am" - update the cron job schedule using cron.update.
"stop notifications" - remove the cron job using cron.remove.
"send notifications to Slack instead" - update the cron job delivery using cron.update.

Commands (optional shortcuts)

Users may also use /watch add X, /watch list, /source enable tmdb, /platform add netflix etc. These are auto-reply commands that work without the AI agent.