agentleFS
TabooHarmonyTabooHarmony/roblox-brain/skills/design/roblox-growth-design/SKILL.md

roblox-growth-design

Skill150 starsChanged 53 days agoMITView on GitHub

Use for Roblox growth: diagnosing stats, discovery, retention, onboarding, experiments, LiveOps, thumbnails, and packaging.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask for dashboard, release, acquisition, and session evidence.

    line 44
  2. Distinctive:

    Write a falsifiable hypothesis: If Y changes because evidence Z, X should move without harming G.

    line 46
  3. Distinctive:

    Prefer one test with a primary metric, counter-metrics, MDE, and decision rule.

    line 47
  4. Distinctive:

    Stop for safety, severe regressions, or invalid instrumentation.

    line 48
  5. Distinctive:

    Place products at points of interest with high foot traffic, and time prompts at the decision moment (e.g., 2x offline earnings at the …

    line 78
  6. Distinctive:

    After shipping, check collateral damage: playtime, D1, D7, FTUE, and in-game economy.

    line 80
  7. Distinctive:

    Use accurate metadata. Avoid dark patterns, deceptive odds, coercive scarcity, and "whale" targeting.

    line 84

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Roblox Growth Design

When to Load

Load for growth, discovery, retention, onboarding, packaging, or LiveOps; route implementation to domain skills.

Quick Reference

Diagnose before prescribing

  1. Ask for dashboard, release, acquisition, and session evidence. Never invent metrics.
  2. Find the narrowest broken transition: impression→play, join→control, action→payoff, return, or purchase.
  3. Write a falsifiable hypothesis: If Y changes because evidence Z, X should move without harming G.
  4. Prefer one test with a primary metric, counter-metrics, MDE, and decision rule.
  5. Stop for safety, severe regressions, or invalid instrumentation; avoid significance peeking.
<!-- temporal: 2026-07 -->

Home diagnosis

  • Low Play-Through Rate (PTR, was QPTR): packaging readability, promise accuracy, or audience mismatch. Also check Experience Detail Page CTR.
  • High First-Play Bounce Rate (negative stat): misleading package/clickbait, join failure, performance, confusing FTUE, or delayed payoff.
  • Low D1: core loop, onboarding, stability, goals, or first payoff.
  • Low D7/D30: progression, variety, social value, LiveOps, endgame, or exhaustion.
  • Low conversion/ARPPU: product fit, value communication, friction, catalog depth, or concentration.
  • Low 7-day spend days / play days per user: consumables, purchase pathways, or event cadence.

These are hypotheses, not one-to-one causes. Segment and test. D1 can move without a release as acquisition mix or social density changes (§3). Per-stat tactics in full.md §3–§6.

Algorithm hotspots (2026)

In this section, (official) marks documented behavior; other ranking claims are unverified.

  • Cold traffic: ads are the cheapest, least-qualified players; Home algo ranks the qualified audience. Launch stats look bad on ad players. Don't panic.
  • Ranking: Roblox docs say benchmark games do not affect ranking (official); the similar-games benchmark is rough context only. An unverified hypothesis adds staging: genre-wide first, then against "experiences with similar players," where better-statted competitors take your players. Use the benchmark to choose comparison cohorts, not as a ranking rule.
  • Ads do not buy Home placement; meaningful stat-improving updates do. 28-day signals (D1, D2–7, D8–28); updates take days to show.
  • Beta mode hides a game from Home (official feature); tune metrics on cheap ad traffic before opening to the algo.
  • The 250 highly-engaged-player requirement (official): all-ages games start 16+ only until 250 highly engaged age-checked plays within 60 days. Definition (official, evolving): account tenure + playtime in your game + a purchase anywhere on Roblox in the last 60 days. 25 = publishing-fee refund threshold (separate). Ads serve 16+ automatically. Home impressions accelerate the count fastest.
  • Experience detail page matters: put your best thumbnails and a gameplay description there; it feeds overall play-through.
  • Launch realism: split ad campaigns desktop/mobile; paid playtime ≈ half of organic; low-CCU stats are noise (don't kill on a 3% D1 readout); algo "biting" shows as thumbnail impressions before CCU; console = cheap long-session niche; multiplayer needs singleplayer fallbacks; feedback forms at decision popups.

Monetization mental model

  • Convert valuable one-time game passes into consumables (repeat-purchase dev products): they pump 7-day spend days per user and solve recurring pain.
  • Products must be must-haves that solve a pain (effort grind, losing progress AFK, FOMO, status). Limiteds with real scarcity outsell unlimited cosmetics.
  • Place products at points of interest with high foot traffic, and time prompts at the decision moment (e.g., 2x offline earnings at the collect point).
  • Give players multiple purchase pathways: HUD shop icon, "+" next to currency, and a cash-shop popup after repeated failed buys.
  • After shipping, check collateral damage: playtime, D1, D7, FTUE, and in-game economy.

Guardrails

Use accurate metadata. Avoid dark patterns, deceptive odds, coercive scarcity, and "whale" targeting. Treat moderation, abuse, accessibility, localization, performance, and economy health as constraints.

Full signal definitions, positioning, FTUE, packaging, social design, monetization mental model, algorithm mechanics, LiveOps, and audit workflow: references/full.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.