Retention strategy designer
Skill ajitta/Game-Engagement-Retention-Skills/skills/retention-strategy-designer
Analyze an app, game, SaaS, content product, interactive narrative, community, or AI product and propose lifecycle retention improvements. Use when the user asks to increase D1, D7, D30, weekly/monthly retention, cohort curves, activation, habit formation, churn, resurrection, onboarding, notification strategy, or return behavior — or when the named deliverable is a multi-day cadence mechanic (battle pass, login rewards, streaks, daily quests). Do NOT trigger when the requested change is a specific in-session scene, reward reveal, staging fix, or a session-length/quit-mid-session problem — even if D1/D7/onboarding is cited as the motivating metric — use interaction-reward-moments and keep the metric as the validation target. Do NOT trigger when moment design is a second deliverable, a moment complaint is paired with churn, or the ask is which moment drives retention — use engagement-retention-advisor. Outputs retention diagnosis, segments, lifecycle loops, experiments, and anti-dark-pattern guardrails.From its SKILL.md
npx -y skills add ajitta/Game-Engagement-Retention-Skills --skill retention-strategy-designerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
6.5 KB, as published. Nobody here has run it
Retention Strategy Designer
Retention is not "getting users addicted." Retention is users returning repeatedly because the product keeps delivering value on a cadence that matches their intent.
See references/retention-playbook.md for frameworks, benchmarks, and experiment design. For per-domain ethical guardrails (fortune/saju, AI companion, narrative, learning, games), read ../engagement-retention-advisor/references/domain-ethics.md (if installed).
Routing
- User asks for retention only (churn, D1/D7/D30, lifecycle, habit, return, cohort, notifications, onboarding), or the named deliverable is a multi-day cadence mechanic (battle pass, daily login rewards, streaks, daily quests) → this skill.
- User asks only for moment-level satisfaction, reward points, dopamine points, scene design, or a session-length / quit-mid-session problem →
interaction-reward-moments. Same even if D7/onboarding is cited only as motivation, as long as the deliverable is a scene fix. - User asks for both deliverables (moment design + retention strategy), the moment-to-return link itself is the question, or a moment complaint is paired with a churn/return problem ("combat feels flat and players churn") →
engagement-retention-advisor. - Ambiguous and the choice materially changes the output → ask one clarifying question.
- When routing away, invoke the target skill; if the Skill tool is unavailable, Read
../interaction-reward-moments/SKILL.mdor../engagement-retention-advisor/SKILL.mdand follow its workflow.
Workflow
-
Define the retention target:
- Return event: the action that proves meaningful return (not a bare app open).
- Usage cycle: daily / weekly / monthly / episodic / event-driven / seasonal — respect the product's natural cadence. Do not force daily retention on a weekly product.
- Lifecycle stage: new / current / power / lapsing / dormant / resurrected.
-
Diagnose leak points:
- D0/D1: onboarding, activation, time-to-value, friction to the first useful result (the "aha moment").
- D2-D7: habit loop, reminder relevance, unfinished goals, value of the second session.
- D8-D30: content depth, progression, personalization, social loops, identity investment.
- Long-term: novelty pipeline, compounding value (archive, records, mastery), community, trust.
-
Propose 3-5 retention interventions:
- Each intervention specifies segment, trigger, user value, implementation, metric, experiment, risk.
- Value loops > push-notification patches. Notify only when it is genuinely useful to the user right now.
- Bind each proposal to a specific return event + time window.
-
Measurement / experiment design:
- Measure by cohort (whole-population averages are contaminated by new inflows).
- Bound the novelty effect: observe for 2-4+ weeks.
- Track guardrail metrics in parallel: notification opt-out rate, uninstalls, complaints, session quality.
Output
Answer in the user's language (Korean input → Korean output); keep the section/table structure below — headings may stay in English or be translated.
## Input interpretation
- Product:
- Core value:
- Recommended return event:
- Natural usage cycle:
- Diagnosed leak points:
## 3-5 retention proposals
| # | Proposal | Target segment | Lifecycle stage | When it fires | User value | Implementation | Core metric | Experiment | Risk |
|---|---|---|---|---|---|---|---|---|---|
## Prioritization
(order by impact × implementation difficulty, with a 1-line rationale each)
## Measurement design
- Return event definition:
- D1 / D7 / D30 (or W1/W4, matched to the product cycle):
- Cohort split criteria:
- Guardrail metrics:
Guardrails
Apply guardrails as design constraints before generating proposals — not as a post-hoc filter.
- Do not optimize opens, session length, or notification clicks alone. Optimize a meaningful return event.
- Do not recommend spam push, fear, guilt, FOMO, fake scarcity, punishing the daily schedule with expiring rewards, hidden friction, or pain-relief monetization.
- Treat staging that periodically reminds users of loss — even truthfully (decaying meters, repeated "before you forget" warnings) — as a fear mechanism.
- Notifications must state a frequency cap (default: recommend ≤1/day) AND be content-gated — the cap is a ceiling, not an entitlement. For fortune/tarot/horoscope, AI companion, and journaling products, recommend only notifications carrying user-derived content (saved-reading follow-up, weekly summary); before proposing notifications in these domains, read
../engagement-retention-advisor/references/domain-ethics.md. No repeated deadline pressure under the guise of "useful info." - No guilt streaks — replace with recovery-friendly progress (streak freeze and recovery credit that are free or earned through use, never sold; flexible goals). Never sell relief from streak loss.
- Separate retention from monetization pressure. Flag when return-driving degrades into spend-driving.
- Manipulative loops retain users but destroy trust. When retention numbers conflict with user value, state it and prioritize user value.
- When a prohibited pattern is the request itself: state the refusal rationale explicitly in the answer body (not a quiet table-cell note), then redesign an ethical alternative that meets the same business goal. No quiet partial compliance.