Bolta.review.digest

Goal

Produce a concise, actionable summary of review activity for human consumption.

V2 Changes

• Agent-aware: References which agents created content, which approved it
• State machine updated: Uses V2 states (inbox, approved, scheduled, published)
• Client tag filtering: Agency users can get per-client digests
• Performance metrics: Includes engagement data if available

Steps

1. Fetch inbox items from period:

   • bolta.list_inbox_items({workspace_id, created_after: period_start, limit: 100})
   • Filter by client_tag if provided

2. Calculate stats:

   • Total items in inbox during period
   • Items approved vs rejected vs still pending
   • Average time-to-approval
   • By agent: posts created, approval rate, average quality

3. Identify top performers:

   • bolta.list_recent_posts({workspace_id, published_after: period_start, limit: 50})
   • bolta.get_post_metrics(post_id) for each
   • Rank by engagement (likes, comments, shares)

4. Generate action items:

   • "X posts still awaiting review (oldest: Y days)"
   • "Agent Z has low approval rate (40%) - review persona"
   • "Client ABC has 5 posts ready to schedule"

5. Format summary:

   Weekly Review Digest (Feb 14-20, 2026)
   
   📊 Stats:
   - 45 posts created (by 3 agents)
   - 32 approved (71% approval rate)
   - 8 rejected (18%)
   - 5 still pending review
   
   🏆 Top Performers:
   1. "How we 10x'd engagement" - 234 likes, 45 comments
   2. "Founder story: Why we built this" - 189 likes, 32 shares
   
   🤖 Agent Performance:
   - HypeMan: 20 posts, 85% approval rate ⭐
   - ContentBot: 15 posts, 60% approval rate ⚠️
   - WeeklyPlanner: 10 posts, 100% approval rate 🎯
   
   ✅ Action Items:
   - 5 posts pending review (oldest: 3 days)
   - ContentBot needs persona review (approval rate dropped)
   - Client Acme Corp: 8 posts ready to schedule
   

Output

Return summary_text (markdown formatted), stats (object), top_performing_posts, action_items.

Agent Types That Use This Skill

• reviewer — Generate digest as part of weekly review job
• analytics — Broader reporting capabilities
• custom — User-defined agents with reporting enabled