social-cli-mcp: Instructions file for Claude Code

v7.0 | flutur-cc | 2026-02-07

GRAFO

S.1 "SurrealDB" :graph @localhost:8000 ns=social db=analytics root/root .schema → src/db/schema.surql .fn → {outreach_status, data_freshness, top_hashtags, daily_analytics, recent_correlations}

S.2 "Supabase" :storage+ai | file storage + Claude Vision labeling .schema → supabase/migrations/001_photos_schema.sql .client → src/db/supabase-client.ts .sync → npx tsx scripts/sync-supabase-photos.ts [import|label|sync|status]

S.3 "Keychain" :credentials service="social-cli-mcp" .api → src/core/credentials.ts | getFromKeychain() setInKeychain() .ok = {Twitter, Instagram, Gmail, Telegram, Anthropic} .setup = {YouTube, Supabase} .missing = {LinkedIn, TikTok}

S.4 "8i8.art" :epk Next.js 14.2→Vercel .config → link-hub/src/lib/artist-config.ts .utm = ?utm_campaign=beach → hero swap .cookies = zero σ₂ send link NOT PDF | PDF only when explicitly requested

ENTITÀ CENTRALI

E.1 "artist_profile:flutur" :nodo-centrale @S.1
  →creates→ E.2
  →performs_at→ venue
E.2 "software_project:jsom" :exit @github.com/alemusica/jsom
  →targets→ {Lovable, Vercel, Prisma, YC}
  =$ 6-7 fig

email →sent_to→ venue
content →taken_at_gig→ gig →performed_at→ venue
post →uses_hashtag→ hashtag
post →belongs_to_pillar→ content_pillar
story_fragment →fragment_inspires→ platform_content
story_fragment →fragment_in_arc→ story_arc
web_research →memory_link→ entity (via researchStore.save)

FLUSSI DATI

M.1 instagramFetcher ← IG API (10 calls/wk budget) → S.1:audience_snapshot
M.2 youtubeFetcher ← YT API OAuth2 → S.1:youtube_snapshot
M.3 correlator ← {M.1, M.2, S.1:epk_analytics} → S.1:analytics_correlation
  | email→views→visit→reply pattern, 3-day window
M.4 insightsArchiver → M.5 → M.6
M.5 editorialIntelligence | corridor IT↔GR 70%
M.6 contentDrafter :interview | NON inventa, chiede
  .styles = {storytelling, minimal, poetic, educational, bridge}
  .guard = "Cosa NON devo inventare?"
M.7 email-guard :safety fail-CLOSED
  .limit = 25/day
  .log → logs/send-log.json (append-only)
  → gmail-sender → S.1:email →sent_to→ venue
M.8 duplicateChecker :auto
  | exact + 70% similarity + topic cooldown 7d
  .track → {analytics/posted-tweets-index.json, analytics/posted-instagram-index.json}
M.9 storyStore :editorial | biographical narrative fragments
  .save → story_fragment → memory_link (entities)
  .query → byTheme, forChannel, unpublished, byEntity
M.10 researchStore :cache | web research persistence
  σ₂ BEFORE WebSearch → researchStore.find(topic) | skip if <7d
  σ₂ AFTER WebSearch → researchStore.save() | ALWAYS
  .query → find(topic), search(query), recent()
M.11 gmailReader :inbox | Gmail API scan + Message-ID dedup
  .scan → outreach_reply (SCHEMAFULL, UNIQUE gmail_message_id)
  .scanSent → email (sent folder scan)
  .getThread → full Gmail thread (sent+reply)
  .findThreadByMessageId → RFC2822 → gmail_thread_id
  .classify → human_reply | auto_reply | bounce
  .persist → outreach_reply + memory_link(venue)
M.12 conversationStore :threads | unified sent+reply thread view
  .get(venue) → OutreachConversation (chronological messages)
  .dashboard() → all conversations, status, action needed
  .backfillThreadIds() → link email records to Gmail threads
  .syncTracking() → tracking.json → email table
  .getGmailThread(venue) → live Gmail API thread
M.13 intelligenceRouter :orchestrator | event→action dispatcher
  .route(event) → IntelligenceBriefing {logistics, cluster, conversation, actions}
  .routeReply(venue, email, type, preview) → auto-briefing on reply
  .formatTelegram(briefing) → Telegram digest with costs + cluster + actions
  .formatConsole(briefing) → full console output
  | AUTO-TRIGGERS: morning-check runs router on every human reply
  | Logistics dept: flight + baggage + accommodation + break-even
  | Cluster dept: nearby venues in same country, contactable count
  | Conversation dept: thread status, suggested next action