OiMy v8.5+ Enhancement Plan

Executive Summary

Research-driven prioritization across 9 products. Three moves that matter most for the next 4 weeks.

Products Researched

Adopt Recommendations

+0.8

Projected NPS Gain

4 wk

Implementation Window

Key Decisions

Proactive re-engagement is the #1 priority. Every competing product (Jetty, Hirey, Pine) has some form of proactive outreach. OiMy currently has none. This is the single biggest lever for NPS and retention in the critical first 7 days.
Pine Voice API is the strongest infrastructure addition. Enables subscription cancellation, appointment booking, and bill negotiation — directly serving the "reduce mental load" value prop for parents. Already has an OpenClaw plugin.
Jetty's conversational check-in pattern is worth adapting. Voice-first, natural language symptom tracking that auto-extracts patterns. Maps directly to OiMy's wellness and symptom-triage skills. The "doctor-ready summary" concept is a high-value add-on.
Tradclaw's skill set has 4 direct overlaps (meal-planner, homework-log, home-maintenance, calendar-briefs) that can validate/improve OiMy's existing scaffolds.
Happenstance, Hirey, Yafafits: skip for now. Professional networking, agent-to-agent matching, and outfit logging don't serve the 30-45yo parent coaching use case well enough to justify integration cost.

Research Verdicts Summary

Product	What It Does	Verdict	Rationale
Jetty Health	AI symptom tracker via voice check-ins, pattern detection, doctor summaries	Adopt	Check-in pattern + auto-insights map perfectly to wellness/symptom-triage skills
Pine Voice	Real phone calls: bill negotiation, subscription cancellation, appointment booking	Adopt	OpenClaw plugin exists. Directly reduces parental mental load. 93% success rate.
Tradclaw	Household skills: meal-planner, homework-log, home-maintenance, calendar-briefs	Adopt	4 skill overlaps for cross-validation. Meal-planner and homework-log particularly strong.
Just-Fucking-Cancel	CC statement analysis → subscription audit → auto-cancellation via browser	Monitor	Great concept but Claude Code-specific. Pine Voice subsumes the cancellation use case.
PrintingPress	Agent-native CLIs: Allrecipes, Wanderlust GOAT, Tesla, table reservations	Monitor	Recipe GOAT and table-reservation interesting for enrichment, not core coaching.
Hirey	Agent-mediated human networking — hiring, dating, cofounders	Skip	Professional/recruiting network. Zero overlap with family coaching.
Happenstance	AI-powered professional network search (Gmail, Calendar, Twitter sync)	Skip	Sales/recruiting tool. Not relevant to parent demographic.
Yafafits	Outfit logger / fashion sync	Skip	Niche fashion tool. No coaching value.
Hermes Skills Hub	88k+ skills across registries (loaded dynamically, couldn't fully scrape)	Monitor	Massive catalog worth mining later for specific OT/speech/elder-care templates.

Proactive Re-engagement System

The #1 strategic gap. OiMy currently has no mechanism to reach out to inactive users. This design is production-ready.

P0 Why This Is Critical +0.4 NPS

Research from Appcues (2026 benchmarks): most apps lose 75% of users in 30 days. The critical window is days 1-7. For a coaching app like OiMy, the value proposition only crystallizes after 3-4 meaningful interactions. If a user doesn't return within 48 hours of onboarding, the probability of natural return drops to <15%.

Every competing product researched (Jetty, Pine, Hirey) has some form of proactive outreach. Jetty does daily voice check-ins. Pine does async task completion notifications. The absence of this in OiMy is the single biggest gap between current state and 9.0+ NPS.

Architecture: Rhythm Engine Extension

OiMy already has a Rhythm Engine with 8am morning brief and 8pm closeout crons. The re-engagement system extends this with a ReEngagementWorker that runs twice daily (10am, 6pm) and evaluates each user against engagement rules.

// ReEngagementWorker — pseudocode
class ReEngagementWorker {
  schedule: "0 10 * * *" | "0 18 * * *"  // 10am, 6pm user-local

  for each family_container:
    last_msg = get_last_user_message()
    onboarding_date = get_onboarding_date()
    profile = get_deep_profile()
    entities = get_entities()  // children, stressors, goals

    segment = classify(last_msg, onboarding_date, profile)
    // segments: NEW_FRESH, NEW_DORMANT, ACTIVE_LAPSED, LONG_DORMANT

    if segment == NEW_DORMANT:    // onboarded, no msg in 48h
      send(template_for(segment, profile, entities))
      log(sent=true, template_id, next_retry=72h)

    if segment == ACTIVE_LAPSED:  // was active, silent 7+ days
      if days_since(last_msg) >= 7 AND cooldown_ok(last_reengagement):
        send(template_for(segment, profile, last_msg_topic, entities))
        log(sent=true, template_id, next_retry=14d)

    if segment == LONG_DORMANT:   // 30+ days
      if reengagement_count < 3:  // max 3 attempts ever
        send(final_attempt_template(profile))
        log(sent=true, template_id, next_retry=never)
      else:
        mark_as_churned()  // stop trying
}
      

User Segments & Rules

Segment	Definition	Trigger	Max Frequency	Stop Condition
NEW_FRESH	Onboarded <48h ago, has sent ≥1 message	—	No outreach (let them come)	—
NEW_DORMANT	Onboarded <7 days ago, no message in 48h+	48h silence post-onboarding	1 msg at 48h, 1 at 72h, 1 at 120h	User replies OR 3 messages sent
ACTIVE_LAPSED	Previously active (3+ sessions), no activity 7+ days	7-day silence	1 msg at 7d, 1 at 14d, 1 at 30d	User replies OR 3 messages sent
LONG_DORMANT	30+ days no activity, was previously active	30-day silence	1 final message	User replies OR 1 message sent → mark churned

Message Templates — NEW_DORMANT (first 7 days)

These use the FRIEND model tone: warm, specific, uses what we know. No marketing language.

Template ND-1: The Curiosity Hook (send at 48h)

"Hey {{parent_name}}, I was thinking about {{child_name}} today — at {{child_age}}, they're probably hitting the phase where [age-appropriate milestone from profile]. Want to talk about what that looks like for your family? No pressure either way 🌱"

Template ND-2: The Gentle Mirror (send at 72h)

"You mentioned {{last_topic}} when we first talked. I found something that might actually help — {{brief_insight_from_kb}}. Thought of you and wanted to share."

Template ND-3: The Practical Nudge (send at 120h)

"{{parent_name}}, it's been a few days. Here's a quick one — {{child_name}}'s {{known_stressor}} might have a new angle worth trying tonight. Want the 30-second version?"

Template ND-4: The Permission Slip (send at 120h if ND-3 ignored)

"Hey — just checking in once more. If this isn't the right time for you, totally fine. I'll be here when you need me. (And hey, even 'what should I make for dinner?' counts as needing me 😄)"

Message Templates — ACTIVE_LAPSED (7+ days dormant)

Template AL-1: The Contextual Recall (send at 7d)

"{{parent_name}}, last time we talked about {{last_topic}}, you said {{specific_detail_from_memory}}. How did that go? I've been thinking about it."

Template AL-2: The Calendar Bridge (send at 14d)

"{{child_name}}'s {{upcoming_event_from_calendar_or_profile}} might be coming up soon. Last time you mentioned it was stressful — want to plan ahead this time?"

Template AL-3: The Honest Check-in (send at 30d)

"It's been a couple weeks. Sometimes the coaching works best when it's a habit, and sometimes life just gets in the way. Either is okay. If there's one thing on your mind today, I'm here."

Message Templates — LONG_DORMANT (30+ days)

Template LD-1: The Goodbye-With-Dignity (send once)

"{{parent_name}}, haven't heard from you in a while — hope you and the family are doing well. I'll stop reaching out, but I'm always one message away. If you ever want to pick up where we left off, everything about {{child_name}}, {{child_name_2}}, and what we worked on is still here. Take care. 🤍"

Data Fields Used by Templates

parent_name — from deep profile layer 2 (identity)
child_name, child_age — from entity extraction
last_topic — from last conversation's skill router classification
specific_detail_from_memory — from Honcho dialectic memory, last session
known_stressor — from deep profile layer 6 (emotional/relational)
age_appropriate_milestone — from coaching KB lookup (Becky Kennedy / development stages)
brief_insight_from_kb — from knowledge_base_coaching.jsonl, matched to last_topic
upcoming_event — from calendar-manage skill data or profile layer 5 (behavioral patterns)

Anti-Spam Guardrails

Quiet hours: No messages between 10pm-7am user-local time. Rhythm Engine respects existing timezone config.
Max frequency: Never more than 1 proactive message per 48h across all segments.
Global cap: Max 7 proactive messages total across a user's lifetime unless they've replied (which resets the counter).
Reply = reset: Any user reply immediately clears all re-engagement state and returns to normal operation.
Unsubscribe: "Stop reaching out" or "don't message me" → sets reengagement_optout=true, permanent until user manually re-enables.
Degrade gracefully: If entity data is missing (no children names), fall back to generic-but-warm templates. Never send a message with {{placeholder}} unfilled.

Implementation Spec

// File: workers/reengagement_worker.py
// Deps: existing rhythm_engine, honcho_client, telegram_bot_api

class ReEngagementWorker:
    cron_schedule = "0 10,18 * * *"  # twice daily
    segments = {
        "NEW_DORMANT": {
            "condition": lambda u: days_since(u.onboarded) < 7 and hours_since(u.last_msg) >= 48,
            "templates": ["ND-1", "ND-2", "ND-3", "ND-4"],
            "intervals_hours": [48, 72, 120, 120],
            "max_sends": 3
        },
        "ACTIVE_LAPSED": {
            "condition": lambda u: u.total_sessions >= 3 and days_since(u.last_msg) >= 7,
            "templates": ["AL-1", "AL-2", "AL-3"],
            "intervals_days": [7, 14, 30],
            "max_sends": 3
        },
        "LONG_DORMANT": {
            "condition": lambda u: days_since(u.last_msg) >= 30 and u.total_sessions >= 1,
            "templates": ["LD-1"],
            "max_sends": 1
        }
    }

    def evaluate(self, user):
        # 1. Check optout
        if user.reengagement_optout: return None
        # 2. Check quiet hours
        if not in_sendable_window(user.timezone): return None
        # 3. Classify segment
        for seg_name, seg_config in self.segments.items():
            if seg_config.condition(user):
                # 4. Check cooldown
                if not cooldown_ok(user, seg_config): return None
                # 5. Pick next template
                template_id = next_template(user, seg_config)
                # 6. Fill template with entity data
                msg = fill_template(template_id, self.get_context(user))
                # 7. Validate no unfilled placeholders
                if has_unfilled_placeholders(msg): msg = fallback_generic(seg_name)
                return msg
        return None

    def get_context(self, user):
        return {
            "parent_name": user.profile.identity.display_name,
            "child_name": user.entities.primary_child.name,
            "child_age": user.entities.primary_child.age_display,
            "last_topic": user.last_session.skill_id,
            "last_detail": self.honcho.get_last_session_key_fact(user.id),
            "known_stressor": user.profile.emotional.top_stressor,
            "upcoming_event": self.calendar.next_event(user.id),
        }
      

Estimated implementation: 3-4 days. Uses existing rhythm_engine cron infrastructure, Honcho for memory, Telegram Bot API for delivery.

New Skills Recommendations

Top 5 skills derived from research. Each selected for direct NPS impact on the 30-45yo parent demographic.

P0 skill: subscription-audit +0.15 NPS

Source: Just-Fucking-Cancel + Pine Voice Assistant

What it does: Helps parents identify, evaluate, and cancel unused subscriptions. Uses Pine Voice for automated cancellation calls. Parents 30-45 average 12+ subscriptions with 3-4 unused ones (stat from JFC repo).

Trigger phrases (for router prompt):

"cancel my subscription"
"what am I paying for"
"audit my bills"
"stop paying for"
"too many subscriptions"
"which services am I subscribed to"
"help me cancel Netflix/Spotify/etc"
"reduce my monthly bills"
"I'm paying for things I don't use"
"subscription overload"

Coaching scaffold principles:

Empathy-first framing: "Subscription creep happens to everyone — it's not negligence, it's bandwidth. Let's clean it up."
Structured audit: Walk through known services category by category (streaming, productivity, fitness, kids' apps).
Decision coaching: Don't just list — ask "when did you last use this?" and "would you miss it if it disappeared?"
Action bias: Offer to handle cancellation via Pine Voice if available, or provide direct cancellation links.
Follow-up: Schedule a 30-day check-in to see if any cancelled services were missed.

Profile data used:

personal-finance skill data (if available)
Entity extraction: known services mentioned in conversations
Profile layer 5 (behavioral patterns): spending habits discussion

P0 skill: health-checkin +0.15 NPS

Source: Jetty Health — conversational voice check-ins + pattern detection

What it does: Daily or on-demand conversational health check-in. Parent describes how they're feeling (physical + mental), and OiMy extracts patterns over time. Weekly summaries highlight trends. Creates "doctor-ready" summaries when asked.

Trigger phrases:

"how am I doing health-wise"
"check in on my health"
"I've been feeling off lately"
"track my symptoms"
"health update"
"I feel tired/stressed/sick"
"what's been going on with my body"
"prepare a summary for my doctor"
"am I getting worse or better"
"daily check-in"

Coaching scaffold principles:

Conversational, not clinical: "How are you feeling today?" not "Rate your pain 1-10." Follow Jetty's natural language approach.
Auto-extract: From free-text, identify: sleep quality, energy, mood, physical symptoms, appetite, stress level. Store as structured data.
Pattern surfacing: After 3+ check-ins, begin noting trends: "You've mentioned low energy 4 times this week — that's new. Want to explore what's different?"
Doctor summary mode: On request, generate a clean timeline: symptoms, duration, severity changes, notable events.
Wellness integration: Cross-reference with wellness-daily and vagal-tune skills for actionable suggestions.

Profile data used:

Honcho memory: past health discussions
Entity extraction: known conditions, medications
Profile layer 6 (emotional): stress baselines
Wellness skill history

P1 skill: homework-deep +0.12 NPS

Source: Tradclaw homework-log skill — photo → log + tracking

What it does: Parent sends photo of child's worksheet/homework. OiMy extracts the content, logs it, identifies struggling areas, and suggests coaching strategies. Builds a homework history that feeds into parent-coach conversations.

Trigger phrases:

"here's the homework"
"take a look at this worksheet"
"what's on the homework today"
"Johnny's math assignment"
"is this homework too hard"
"help me understand what they're learning"
"homework log"
"what did they get wrong"
"tutor mode"
"worksheet photo"

Coaching scaffold principles:

Parent as coach, not answer-giver: "Here's what they're working on, and here's how to help them figure it out — not the answer itself."
Difficulty assessment: Match worksheet content against grade-level expectations. Flag if content seems advanced or if child appears to be struggling.
Pattern tracking: Over time, identify subject areas where child consistently needs help. Feed into grade-monitor and education skills.
Time management coaching: Estimate completion time and suggest scheduling.
Celebrate wins: When homework shows improvement, proactively note it.

Profile data used:

Entity extraction: child names, ages, grades, schools
Education skill history
Grade-monitor data
Profile layer 5: child behavioral patterns

P1 skill: appointment-negotiator +0.10 NPS

Source: Pine Voice API — automated phone calls for booking and negotiation

What it does: Helps parents book, reschedule, or negotiate appointments (doctor, dentist, daycare, contractor) via automated phone calls. Handles hold times, IVR menus, and callback scheduling.

Trigger phrases:

"book a dentist appointment"
"call the doctor's office"
"reschedule my appointment"
"find a pediatrician"
"call the plumber"
"schedule a consultation"
"book daycare interview"
"call and ask about pricing"
"set up a meeting with the school"
"negotiate with the landlord"

Coaching scaffold principles:

Context gathering: Before calling, confirm: who to call, what outcome is desired, constraints (time windows, budget).
Preparation: Brief the "caller" (Pine Voice) with family context, insurance info if relevant, preferred times.
Escalation awareness: If the call requires human judgment (e.g., medical discussion), flag for parent to handle personally.
Follow-through: After successful booking, add to calendar and set reminder 24h before.

Profile data used:

Entity extraction: healthcare providers, known contacts
Calendar-manage data
Appointments skill history
Healthcare skill data: insurance info, preferred providers

P1 skill: school-triage +0.08 NPS

Source: Tradclaw school-triage skill — forms, theme days, notices

What it does: Parent forwards school emails, photos of permission slips, or notes from the teacher. OiMy extracts action items, deadlines, and schedules them. Handles the "I forgot it was picture day" problem.

Trigger phrases:

"school email"
"permission slip"
"what does the school need"
"picture day"
"theme day"
"field trip form"
"teacher's note"
"school notice"
"what's due for school"
"back to school checklist"

Coaching scaffold principles:

Extract and act: From any school communication, pull: deadline, action required, child name, cost (if any).
Priority sort: Distinguish between "needs a response today" and "informational."
Calendar integration: Auto-add deadlines and events to calendar-manage skill.
Delegation coaching: "This permission slip needs signing — want me to remind you tonight when you're home?"

Profile data used:

Entity extraction: children, schools, teacher names
School skill data
Calendar-manage integration
Homework-check and grade-monitor cross-reference

Expert RAG Knowledge Base Gaps

Three archetype skills (OT, Speech, Elder Care) currently return hardcoded templates. These need real knowledge bases.

P0 occupational-therapy Archetype +0.10 NPS

Current state: DeepSeek path returns hardcoded "consult a professional" response. No real OT knowledge embedded.

Gap analysis — what's missing:

Sensory processing framework: Understanding sensory seeking vs. sensory avoiding behaviors in children 3-12. Key concepts: proprioceptive input, vestibular activities, tactile sensitivities.
Fine motor milestones: Pencil grip development, scissor skills, buttoning/zipping timelines. When to worry vs. when it's normal variation.
Self-regulation toolkit: Heavy work activities, calm-down routines, transition strategies. Specific activities organized by age and setting (home vs. school).
ADHD/autism OT overlap: Executive function coaching, visual schedules, task breakdown strategies.
Red flags: When something is developmentally concerning enough to warrant a formal OT evaluation vs. home support.

Recommended knowledge sources:

Otter Jenny (YouTube) — pediatric OT content, 200k+ subscribers, practical activities
The OT Toolbox (theottoolbox.com) — structured activities by skill area
AOTA (American Occupational Therapy Association) — developmental milestone references
Specific Instagram OT accounts for parent-coaching frameworks

Integration approach:

// Add to knowledge_base_coaching.jsonl
{
  "archetype": "occupational-therapist",
  "sources": [
    {"type": "youtube", "channels": ["otter-jenny", "sensory-processing-101"]},
    {"type": "web", "urls": ["theottoolbox.com/activities"]},
    {"type": "instagram", "handles": ["@pediatricot", "@sensorysmartmom"]}
  ],
  "scaffold_principles": [
    "Always differentiate 'needs evaluation' from 'needs support'",
    "Provide specific activities with setup instructions, not abstract advice",
    "Organize by: sensory, fine motor, self-regulation, daily living",
    "Age-tier all recommendations",
    "Flag safety concerns (choking hazards, equipment needs)"
  ]
}
      

P0 speech-therapy Archetype +0.10 NPS

Current state: DeepSeek path returns generic advice. No SLP-specific knowledge or assessment frameworks.

Gap analysis — what's missing:

Language milestone framework: By age 1: 1-3 words. By age 2: 50+ words, 2-word phrases. By age 3: 3-4 word sentences, 75% intelligible. When to refer.
Articulation norms: Which sounds develop when (/r/ is last, around 6-7). When a lisp is normal vs. concerning.
Home practice activities: Specific, play-based language stimulation techniques. Narrating, expanding, recasting. NOT drill-based.
Bilingual considerations: Code-switching is normal. Delay in bilingual children is a myth. How to support dual-language development.
Stuttering guidance: Normal disfluency (ages 2-5) vs. true stuttering. Parent responses that help vs. hurt.

Recommended knowledge sources:

Toddler Talk (YouTube/Instagram) — SLP content for parents
Speech Sisters (speechsisters.com) — milestone frameworks
ASHA (American Speech-Language-Hearing Association) — evidence-based norms
Laura Mize (talktomekids.com) — practical parent coaching

P1 elder-care Archetype +0.08 NPS

Current state: Hardcoded template with generic "consult your doctor" responses. This is a growing need as OiMy's user base (30-45yo) increasingly has aging parents.

Gap analysis — what's missing:

Caregiver burnout framework: Signs, stages, interventions. The "sandwich generation" problem (caring for kids + aging parents simultaneously).
Cognitive decline indicators: Normal aging vs. MCI vs. dementia. What's alarming vs. what's expected.
Care navigation: Medicare/Medicaid basics, assisted living vs. home care decision framework, power of attorney, healthcare proxy.
Difficult conversations: How to talk about driving cessation, memory concerns, moving to care, end-of-life wishes.
Safety modifications: Home safety checklists, fall prevention, medication management systems.

Recommended knowledge sources:

AARP caregiving resources
Alzheimer's Association (alz.org) — early warning signs, communication guides
Family Caregiver Alliance (caregiver.org) — burnout assessment tools
YouTube: Teepa Snow's Positive Approach to Senior Care — dementia-specific coaching

Strategic note: The sandwich generation problem is the highest-growing pain point for OiMy's core demographic. Investing in elder-care depth now positions OiMy ahead of competitors who only focus on parenting. This is a 2-3 year competitive moat.

Implementation: Unified Expert RAG Pipeline

Instead of building each archetype separately, create a unified ingestion pipeline:

// Pipeline architecture
1. Source ingestion: YouTube transcripts (whisper) + web scraping + IG caption extraction
2. Chunking: 500-token chunks with archetype metadata tags
3. Embedding: OpenAI text-embedding-3-small → vector store
4. Retrieval: At inference time, archetype-aware RAG with hybrid search (keyword + vector)
5. Injection: Retrieved chunks injected into DeepSeek/Kimi system prompt with archetype scaffold

// Each archetype gets:
- 50-80 curated chunks (not just raw transcripts)
- Scaffold principles (5 per archetype)
- Confidence thresholds (if RAG returns low-confidence, fall back to "consult a professional")
- Red flag detection (specific phrases that trigger "this needs a real doctor" escalation)
      

Infrastructure Verdicts

Quick verdicts on infrastructure/tooling items from the research. Skip/Monitor/Adopt with reasoning.

Pine Voice API — Adopt

What: Real phone call API. AI agent dials, navigates IVR, negotiates, returns transcripts. 93% success rate across 50k+ users.

Why adopt: Already has an OpenClaw plugin (openclaw-pine-voice). Directly serves the subscription-audit and appointment-negotiator skills. Pine Assistant handles multi-step workflows (research → strategy → call → follow-up). This is the single biggest functional expansion that maps to OiMy's "reduce mental load" value prop.

Cost: Pine AI Pro subscription (monthly credits). Per-call cost is absorbed by Pine's plan structure.

Risk: English only currently. Calls limited to US, UK, AU, NZ, SG, IE, HK (+1, +44, +61, +64, +65, +353, +852). Most OiMy users are English-speaking with US phone numbers, so this is acceptable for v1.

Timeline: 2-3 days to install plugin + configure auth + wire into appointment-negotiator and subscription-audit skills.

Browser Automation (CUA/browser-use) — Monitor

What: Computer-use agents that can navigate web UIs — filling forms, clicking buttons, handling complex web flows.

Why monitor: The subscription cancellation use case (from Just-Fucking-Cancel) uses browser automation. However, Pine Voice handles the hardest part (phone calls) without browser automation. For web-based cancellations (Netflix, Spotify), direct deep links to cancellation pages are more reliable than browser automation. Monitor for 6 months — when browser-use becomes more stable and cheaper, it could replace Pine for web-only tasks.

Verdict: Don't build on it yet. Pine Voice covers the voice-call side; direct API/URL approaches cover the web side. Browser automation is the bridge between them that isn't needed right now.

Hirey / Happenstance — Skip

What: Agent-mediated professional networking and people search.

Why skip: Zero overlap with OiMy's family coaching use case. Hirey is for hiring/dating/cofounder matching. Happenstance is for sales/recruiting network search. Neither serves parents trying to manage family logistics.

Hermes Skills Hub — Monitor

What: 88k+ skills across registries. Dynamically loaded catalog with search.

Why monitor: Interesting as a potential source for OT/speech/elder-care skill templates, but the catalog wasn't fully scrapeable (it loads dynamically). Worth revisiting when searching for specific expert archetypes. The skill format (SKILL.md + YAML scaffold) is compatible with OiMy's scaffold approach.

PrintingPress CLIs — Monitor

What: Agent-native CLIs for Allrecipes, Wanderlust GOAT (travel), Tesla, table reservations, and more.

Assessment per CLI:

Recipe GOAT: Monitor — Enriches meal-planning skill with USDA nutrition data and recipe ranking. Not critical (current meal-planning works fine), but would add nutritional depth.
Table Reservation GOAT: Monitor — Cross-books OpenTable + Tock + Resy. Interesting for date-night planning but low priority.
Allrecipes CLI: Skip — OiMy's meal-planning already generates recipes via model. Adding a recipe search CLI adds complexity without proportional value.
Wanderlust GOAT: Skip — Travel enrichment is already covered by travel-planning and travel-itinerary skills.
Tesla CLI: Skip — Niche device control. Not relevant to family coaching.

Yafafits — Skip

What: Outfit logger / fashion sync tool.

Why skip: Pure fashion/lifestyle tool with no coaching dimension. Doesn't serve the family companion mission. The site barely loads (just shows "Outfit Logger" with no features visible).

Implementation Roadmap

4-week phased plan. Each phase is independently shippable. Priority order maximizes early NPS impact.

Week 1: Re-engagement Foundation

P0 +0.4 NPS (projected)

Task	Days	Output
Build ReEngagementWorker class	1-2	workers/reengagement_worker.py with segment classification logic
Write all 8 message templates with entity resolution	2-3	templates/reengagement/ directory with ND-1..4, AL-1..3, LD-1
Wire into Rhythm Engine cron (10am/6pm)	3	Cron job registered in rhythm_engine config
Add anti-spam guardrails (quiet hours, cooldown, opt-out)	4	Guardrail module + unit tests
Test with 5 family containers (manual trigger)	5	Validated message quality, no unfilled placeholders

Week 2: Expert RAG + Pine Voice Integration

P0 +0.25 NPS (projected)

Task	Days	Output
Ingest OT knowledge sources (YouTube transcripts, web content)	6-7	50+ curated chunks in knowledge_base_coaching.jsonl for OT archetype
Ingest speech-therapy knowledge sources	7-8	50+ curated chunks for SLP archetype
Install + configure Pine Voice OpenClaw plugin	8	Pine Voice tools available, auth configured
Build subscription-audit skill scaffold	9	skills/subscription-audit/YAML scaffold + trigger phrases
Build appointment-negotiator skill scaffold	9-10	skills/appointment-negotiator/YAML scaffold + Pine Voice integration

Week 3: New Skills + Elder Care RAG

P1 +0.20 NPS (projected)

Task	Days	Output
Build health-checkin skill (Jetty-inspired)	11-12	skills/health-checkin/ with conversational check-in flow
Build homework-deep skill (Tradclaw-inspired)	12-13	skills/homework-deep/ with photo → extraction → log flow
Build school-triage skill	13	skills/school-triage/ with email/notice → action items flow
Ingest elder-care knowledge sources	14-15	50+ curated chunks for elder-care archetype
Update router prompt with new trigger phrases (all 5 skills)	15	Updated router prompt in triage config

Week 4: Validation + NPS Measurement

Task	Days	Output
Deploy re-engagement to all 38 containers	16	Production deployment with monitoring
Monitor re-engagement reply rates (target: 20%+ reply rate)	16-20	Daily stats: sent, replied, opted-out
A/B test message templates (ND-1 vs ND-2 variants)	17-20	Template performance data
NPS survey round (post-enhancement)	20	New NPS score comparison vs 8.4 baseline
Bug fixes + template tuning based on user feedback	18-20	Refined templates, fixed edge cases

Phased Timeline (4 Weeks)

Initiative	W1	W2	W3	W4	NPS Impact
Re-engagement System					+0.4
Expert RAG (OT/Speech)					+0.2
Pine Voice Integration					+0.1
New Skills (5)					+0.45
Elder Care RAG					+0.08
Validation + NPS Survey					measurement

Total projected NPS impact: +0.8 → target 9.2. This assumes re-engagement alone recovers 15-20% of dormant users (industry average for well-targeted re-engagement is 10-30%). The skill additions add depth to existing coaching conversations; the expert RAG fixes eliminate the most frustrating dead-end responses.

OiMy Enhancement Plan v8.5+

Executive Summary

Key Decisions

Research Verdicts Summary

Proactive Re-engagement System

P0 Why This Is Critical +0.4 NPS

Architecture: Rhythm Engine Extension

User Segments & Rules

Message Templates — NEW_DORMANT (first 7 days)

Message Templates — ACTIVE_LAPSED (7+ days dormant)

Message Templates — LONG_DORMANT (30+ days)

Data Fields Used by Templates

Anti-Spam Guardrails

Implementation Spec

New Skills Recommendations

P0 skill: subscription-audit +0.15 NPS

P0 skill: health-checkin +0.15 NPS

P1 skill: homework-deep +0.12 NPS

P1 skill: appointment-negotiator +0.10 NPS

P1 skill: school-triage +0.08 NPS

Expert RAG Knowledge Base Gaps

P0 occupational-therapy Archetype +0.10 NPS

P0 speech-therapy Archetype +0.10 NPS

P1 elder-care Archetype +0.08 NPS

Implementation: Unified Expert RAG Pipeline

Infrastructure Verdicts

Pine Voice API — Adopt

Browser Automation (CUA/browser-use) — Monitor

Hirey / Happenstance — Skip

Hermes Skills Hub — Monitor

PrintingPress CLIs — Monitor

Yafafits — Skip

Implementation Roadmap

Week 1: Re-engagement Foundation

Week 2: Expert RAG + Pine Voice Integration

Week 3: New Skills + Elder Care RAG

Week 4: Validation + NPS Measurement

Phased Timeline (4 Weeks)