| name | setup |
| description | First-run configuration for the movies-for-ai-agents plugin. Use when the user says "set up movies-for-ai-agents", "configure movies", "first run", "add my TMDB key", "set up my movie profile", or when any other movies-for-ai-agents skill calls an MCP tool that reports "TMDB API key not configured". Captures the TMDB API key and, optionally, a brief taste-priming pass (favorite genres, directors, a few favorite movies) so later recommendations work out of the gate. |
setup
You are setting up the movies-for-ai-agents plugin for the user. This is a one-time flow that writes to user-global storage at ~/.claude/data/movies-for-ai-agents/ (follows the user across every Claude Code project).
Steps
-
Check current status. Call the config_status MCP tool. If configured: true, tell the user it's already set up, show the summary, and ask if they want to re-prime their taste profile. Otherwise proceed.
-
Explain briefly (1–2 sentences) what the plugin does and that a free TMDB API key is required. Link the URL. Mention that IMDb ratings and richer discovery filters come from imdbapi.dev automatically — no extra key needed.
-
Ask for the TMDB API key. Accept either a TMDB v3 API key or a v4 read access token. When the user provides it, call config_set_api_key with apiKey. The tool performs a live smoke-test — if it errors, surface the error and ask them to retry.
-
Offer taste-priming (optional). Say something like "Want to spend 30 seconds priming your taste profile? I'll ask 3 quick questions and recommendations will work better from the start." If yes, ask:
- 3 genres you love and 1–2 you avoid. Call
preferences_set with likedGenres, dislikedGenres.
- 2–3 directors whose work you'd watch sight unseen. Call
preferences_set with favoriteDirectors.
- 3–5 movies that represent your taste well. For each title they give, call
movies_search to resolve it to a TMDB ID, confirm the match if ambiguous, then call preferences_set with favoriteMovies: [ids...].
If they decline, skip priming — preferences fill in naturally as they watch things over time (the movies-journal skill handles that).
-
Confirm. Call config_status again and show the result. Tell them they can now:
- "Find me that 90s sci-fi thriller…" → search
- "What's trending this week?" → discover
- "What should I watch tonight?" → recommend
- "I just watched X, loved it" → auto-logs to history
- "Add X to my halloween list" → custom named lists, organized by theme
- When you start a suggested film, it's tracked as active — next session I'll ask how it went
Notes
- The TMDB key is stored in
~/.claude/data/movies-for-ai-agents/config.json (chmod 600). Alternatively, the user can set TMDB_API_KEY in their environment — env var takes precedence over the stored value.
- Never echo the TMDB key back in your response after saving it.
- If the user has
TMDB_API_KEY already set, skip the key prompt and jump straight to taste-priming.