Bold Little Oracle – Automation Plan

Created: April 3, 2026 Goal: Automate 99% of BLO growth operations using the Cambium agent roster Target: 0 to 10K downloads with $0 ad spend Agent roster: Growth Orchestrator, Vega (marketing), Tesla (research), Shade (UX/reviews), Cipher (QA)

Agent Roster Quick Reference

1. Content Automation

1.1 Daily Social Content Pipeline

The 30-day content calendar defines the weekly rhythm. Agents execute it, not improvise it.

Card-of-the-Day Post Generation

How it works:

1. Tesla picks today’s card using a rotation algorithm across the 72 cards x 2 tones, weighted by: - Theme rotation (Action -> Boundaries -> Self -> Clarity cycle) - Day-of-week format from the content calendar (Monday = theme spotlight, Friday = Sassy Friday, Sunday = weekly pull, etc.) - Cards not yet featured get priority
2. Tesla drafts the caption using the content calendar’s format specification for that day
3. Vega selects the hashtag set (A, B, or C per calendar rules), writes the CTA based on app status phase (in review / live / conversion push), and assembles the final post package
4. Post package is written to content/posts/YYYY-MM-DD-{platform}.md with all metadata
5. During approval period (days 1-7): post is sent to Discord #blo-growth for human approval with a 4-hour window before scheduled post time
6. After approval period: post is auto-queued to the scheduling tool (Buffer, Later, or native scheduling)

Platform-specific rules: - TikTok: video-first, trending sound suggestions included, Set A hashtags for reach - Instagram Reels: same video with platform-specific aspect ratio, Set B hashtags for engagement - Instagram Stories (Tuesdays): poll/quiz slides generated as image assets - Instagram Carousels: Canva template populated with card content

Card rotation tracking: - content/card-rotation-log.yaml tracks which cards have been featured, on which platform, in which tone - Tesla checks this before selecting the next card to ensure full deck coverage over ~36 days (72 cards / 2 per day)

1.2 Weekly Email Automation (Beehiiv)

How it works:

1. Tesla selects a card that was NOT used in social content that week (to keep email subscribers getting exclusive content)
2. Tesla writes a short intro paragraph connecting the card to the week ahead
3. Vega formats the email using the Beehiiv template, including: - Card name as heading - Gentle reading (full text) - Sassy reading (full text) - CTA: “Pull your own card” linking to App Store - Footer: social links, unsubscribe
4. Email is scheduled for Monday 9am delivery via Beehiiv API
5. Growth Orchestrator tracks open rate, click rate, unsubscribe rate after 48 hours

1.3 Content Calendar Execution

How it works:

1. Growth Orchestrator reads docs/content-calendar-30day.md and the current date
2. Identifies today’s required content: platform, format, theme, concept
3. Dispatches to Tesla (content ideation) and Vega (asset production) with the day’s spec
4. Tracks completion in content/execution-log.yaml
5. If a post is missed (agent failure, approval timeout), Growth Orchestrator flags it in the 8pm report and reschedules for next available slot

After Day 30: Growth Orchestrator generates the next 30-day calendar based on: - Performance data from month 1 (which formats performed best) - Remaining un-featured cards - Any new formats Tesla has proposed through AutoResearch

1.4 Approval Workflow

Graduated autonomy model:

Approval mechanism: - Post preview sent to Discord #blo-growth with reaction buttons - Thumbs up = approved, post queues immediately - Thumbs down = rejected, Vega revises and resubmits - No response within 4 hours = auto-post (after day 7) - No response within 4 hours = hold and flag (days 1-7)

What always needs approval (the hard gates): - First post in any new format never used before - Any post mentioning pricing changes - Responses to controversy or negative PR - Paid promotion or ad spend (even $1)

2. ASO and Store Optimization

2.1 Keyword Monitoring

Tracked keyword categories: - Primary: oracle cards, oracle card app, daily oracle, tarot app - Differentiator: sassy oracle, oracle two tones, honest oracle cards - Adjacent: daily card pull, spiritual guidance app, self reflection app - Competitor brand: (tracked but not targeted in copy)

How it works:

1. Vega queries App Store Connect API for impression data by keyword (Search Ads data or third-party ASO tool API)
2. Tesla compares this week vs. last week, calculates rank changes
3. Report format: Keyword Report -- Week of [date] UP: "oracle cards" #12 -> #9 (+3) DOWN: "tarot app" #45 -> #52 (-7) NEW: "sassy oracle" -- first appearance at #28
4. If any primary keyword drops more than 10 positions, Growth Orchestrator triggers the ASO optimization workflow

2.2 A/B Test Listing Copy

What gets tested:

How it works:

1. Tesla generates 2 variant copies based on keyword data and competitor analysis
2. Vega formats them for App Store Connect (character limits, formatting rules)
3. Package sent to Discord #blo-growth for approval with both variants shown side by side
4. Approved variant deployed via App Store Connect API
5. Growth Orchestrator measures conversion rate delta after 7 days
6. Winner becomes the new baseline; loser informs the next hypothesis

2.3 Competitor Monitoring

Monitored competitors (top 5 oracle/tarot apps by downloads): - Track: pricing changes, new features, rating changes, keyword changes, screenshot updates - Tesla maintains a competitor profile file per app in research/competitors/ - Any significant move triggers a Growth Orchestrator alert + recommended response

How it works:

1. Tesla checks each competitor’s App Store listing for changes (description, screenshots, price, version notes)
2. Tesla checks their social media for new campaigns or viral content
3. Significant changes logged to research/competitors/changelog.yaml
4. If a competitor launches a feature BLO doesn’t have, Tesla adds it to the feature ideation backlog with a priority score
5. If a competitor drops price, Growth Orchestrator flags for strategic pricing discussion (human decision)

2.4 Review Monitoring and Response

How it works:

1. Shade pulls new reviews from App Store Connect API
2. Each review is classified: - Positive (4-5 stars): Draft a brief, genuine thank-you response. Reference something specific from their review. - Constructive (3 stars): Draft a response acknowledging the feedback, mentioning if it’s on the roadmap. - Negative (1-2 stars): Draft a response, flag in Discord with the label NEEDS-HUMAN, include sentiment analysis and suggested action. - Bug report in review: Flag for Cipher to investigate. Draft response promising a fix timeline.
3. Shade also extracts recurring themes from reviews and logs them to research/review-themes.yaml
4. If 3+ reviews mention the same issue in a 7-day window, Growth Orchestrator creates a priority issue

Response tone guidelines (Shade follows these): - Never defensive, never corporate - Match the user’s energy – casual reviewer gets casual response, detailed reviewer gets detailed response - Always mention something specific from their review (no canned responses) - If they mention the Sassy tone positively, lean into it slightly in the response

3. Growth Experiments

3.1 AutoResearch Loops

The AutoResearch cycle:

1. HYPOTHESIZE (Tesla)
   "If we post card-of-the-day content at 7pm instead of 10am,
    engagement will increase by 15% because our audience is most
    active in the evening."

2. DESIGN (Tesla)
   Method: A/B by day. Mon/Wed/Fri = 10am, Tue/Thu = 7pm.
   Duration: 2 weeks. Metric: engagement rate per post.
   Minimum sample: 5 posts per time slot.

3. EXECUTE (Vega)
   Adjust posting schedule for the test period.
   Ensure content quality is comparable across both groups.

4. MEASURE (Growth Orchestrator)
   After 2 weeks, compare engagement rates.
   Statistical significance check (is the sample large enough?).

5. DECIDE (Growth Orchestrator)
   Keep: adopt the winning variant permanently.
   Discard: revert and log the learning.
   Iterate: refine the hypothesis and run again.

6. LOG (Growth Orchestrator)
   Write to research/experiments/YYYY-MM-DD-{name}.yaml
   Include: hypothesis, method, data, result, decision, learnings.

Experiment categories (Tesla draws from these): - Posting time optimization - Caption length (short punchy vs. storytelling) - Hashtag set performance - CTA wording variants - Card theme engagement (which of the 4 themes resonates most) - Gentle vs. Sassy ratio in posts - Video length (15s vs. 30s vs. 45s) - Thumbnail/cover image styles

3.2 Pricing Experiments

Pricing test plan:

How it works:

1. Tesla proposes the test when baseline data is sufficient
2. Growth Orchestrator sends pricing change proposal to Discord #blo-growth with projected impact
3. Human approves or modifies
4. Price changed via App Store Connect
5. Growth Orchestrator tracks conversion rate and revenue daily during test
6. After minimum sample reached, Growth Orchestrator calculates RPM for each price point
7. Recommendation posted to Discord with full data table

What we’re optimizing: Revenue per 1000 impressions (RPM), not just conversion rate. A lower price might convert better but earn less per impression.

3.3 Referral Program Optimization

Referral mechanics to test: 1. Share a card reading (screenshot with App Store link) – frictionless, no incentive needed 2. “Pull a card for a friend” – in-app feature that generates a shareable card image 3. Unlock a bonus card for referring 3 friends (requires app code change – flag for human)

Tesla generates hypotheses about which mechanic will work best. Vega creates the social prompts (“share your card pull”). Growth Orchestrator tracks referral-attributed downloads via campaign links.

3.4 Content A/B Testing

Dimensions tracked per post:

Weekly analysis (Tesla): - Rank all posts by engagement rate - Identify which dimension values correlate with top performance - Recommend next week’s content mix based on findings - Flag any format that consistently underperforms for elimination

4. Metrics and Reporting

4.1 Daily Metrics Collection

Data sources and metrics:

Daily metrics file format:

date: 2026-04-15
app_store:
  downloads: 47
  impressions: 1230
  conversion_rate: 0.038
  proceeds_eur: 14.97
  rating_average: 4.7
  rating_count: 23
  new_reviews: 2
social:
  tiktok:
    views: 3400
    followers: 892
    new_followers: 34
  instagram:
    reach: 2100
    followers: 645
    new_followers: 21
email:
  subscribers: 312
  open_rate: 0.43
website:
  visits: 187
  unique_visitors: 156

4.2 Daily Discord Report

Report format:

BLO Daily Report -- April 15, 2026

DOWNLOADS: 47 today (312 total) -- target: 2000 by week 4
REVENUE:   EUR 14.97 today (EUR 89.82 total)
RATING:    4.7 (23 ratings) -- target: 4.5

SOCIAL
  TikTok:    3,400 views | +34 followers (892 total)
  Instagram: 2,100 reach | +21 followers (645 total)
  Top post:  "Sassy Friday #2" -- 1,200 views, 89 saves

EMAIL: 312 subscribers | 43% open rate

SIGNALS
  Positive: Save rate up 15% this week -- Sassy content driving saves
  Watch:    Conversion rate dipped to 3.8% -- below 4% target
  Action:   Vega testing new screenshot order tomorrow

EXPERIMENTS RUNNING
  1. Post time test (7pm vs 10am) -- day 4 of 14
  2. Caption length test -- day 2 of 7

4.3 Weekly Deep Dive

Weekly report sections:

1. Week summary – downloads, revenue, rating, social growth (with week-over-week deltas)
2. Funnel analysis – impressions -> page views -> downloads -> premium conversion, with conversion rates at each step
3. Content performance – every post ranked by engagement, with top 3 and bottom 3 called out
4. Keyword movement – ASO keyword ranking changes
5. Review digest – new reviews, sentiment trend, recurring themes
6. Competitor moves – anything notable from the competitive landscape
7. Experiment results – any experiments that concluded this week, with data and decisions
8. Recommendations – Tesla’s top 3 suggested actions for next week, ranked by expected impact
9. Goal tracker – progress toward Phase 1 goals (2000 downloads, 200 DAU, 4.5 rating, 50 reviews)

4.4 Alert Triggers

Alert definitions:

5. Infrastructure

5.1 Cloudflare Pages Auto-Deploy

How it works:

1. GitHub Actions workflow triggers on push to main
2. Builds static site and deploys to Cloudflare Pages (project: bold-little-oracle, account: bc0c90d4bc8c0d8b4b247f42c57aa65b)
3. Cipher runs post-deploy checks: - HTTP 200 on boldlittleoracle.unicorn.land - Screenshot comparison vs. previous deploy (visual regression) - All critical links working (App Store link, social links)
4. If any check fails, Cipher posts to Discord #app with the failure details

5.2 EAS Build Automation

How it works:

1. GitHub Actions detects code changes in the app directory
2. Triggers EAS Build for iOS and Android
3. Cipher monitors the build queue
4. On success: posts build artifact links to Discord #app
5. On failure: Cipher reads build logs, identifies the error, posts diagnostic with suggested fix
6. App Store submission is always a manual human action (too high-stakes to automate)

5.3 Beehiiv Integration

API integration points: - GET /publications/{id}/subscriptions – subscriber count and growth - GET /publications/{id}/posts/{id}/stats – open rate, click rate, per-email metrics - POST /publications/{id}/posts – create draft email (Vega uses this for Card of the Week) - POST /publications/{id}/posts/{id}/publish – schedule email for send

5.4 App Store Connect API

API integration points: - Sales and Trends API: downloads, proceeds, units by source type - Ratings and Reviews API: new reviews, rating distribution, average rating - App Analytics API: impressions, product page views, conversion rate, sessions - App metadata API (for ASO changes): update subtitle, promotional text, keywords

Authentication: App Store Connect API key (JWT). Stored as a Cambium secret, never in repo.

6. What Stays Human (the 1%)

These items are explicitly excluded from automation. The agents will surface information and recommendations, but a human makes the call.

How agents request human input: 1. Post to Discord #blo-growth with the NEEDS-HUMAN label 2. Include: what they need decided, their recommendation, the data behind it, and a deadline 3. If no response within 24 hours, post a reminder 4. If no response within 48 hours, use the conservative default (don’t post, don’t change, don’t spend)

7. Implementation Phases

Week 1: Content Engine + Daily Reports

Goal: Social content flows automatically; daily metrics are visible.

Week 1 success criteria: - Daily Discord report posting reliably at 8pm - First 5+ social posts published with approval workflow working - Metrics collection running without gaps - Card rotation log tracking which cards have been featured

Week 2: ASO + Review Pipeline

Goal: App Store optimization is monitored; reviews are handled.

Week 2 success criteria: - All new reviews detected within 24 hours - Response drafts posted to Discord for every review - First keyword ranking report delivered - Competitor baseline established

Week 3: Growth Experiments + AutoResearch

Goal: Experiment infrastructure running; first experiments producing data.

Week 3 success criteria: - 2 experiments actively running with data being collected - Content performance ranking identifies clear winners - AutoResearch loop has completed at least 1 full cycle (hypothesis -> execute -> measure -> decide) - Next month’s content calendar drafted

Week 4: Full Autonomous Operation

Goal: All systems running. Human involvement is approval gates only.

Week 4 success criteria: - Social content posting daily without human intervention (except new formats) - Reviews responded to within 24 hours (positive auto, negative human-approved) - Daily and weekly reports posting reliably - Experiments running and producing decisions - All alerts configured and tested - Human involvement is under 15 minutes per day

Appendix A: File Structure

bold-little-oracle/
  docs/
    automation-plan.md          # This document
    content-calendar-30day.md   # Month 1 content calendar
    email-welcome-sequence.md   # Beehiiv welcome series
    marketing-assets.md         # Outreach templates, Product Hunt draft
  content/
    posts/                      # Generated social posts (YYYY-MM-DD-{platform}.md)
    card-rotation-log.yaml      # Which cards have been featured where
    execution-log.yaml          # Content calendar execution tracking
  metrics/
    daily/                      # Daily metrics files (YYYY-MM-DD.yaml)
    weekly/                     # Weekly deep dive reports
  research/
    competitors/                # Competitor profile files
    experiments/                # Experiment logs (YYYY-MM-DD-{name}.yaml)
    review-themes.yaml          # Recurring review themes

Appendix B: Cron Schedule Summary

Appendix C: API Keys and Secrets Required

All stored as Cambium deployment secrets, never committed to the repository.