À propos
La compétence `grow-app` est un cadre guidé pour transformer des applications avec une faible rétention d'utilisateurs en produits dotés de solides boucles d'habitude et de métriques d'activation fiables. Elle orchestre huit compétences spécifiques par phases, en documentant de manière interactive les décisions dans des fichiers projet pour maintenir la progression entre les sessions. Utilisez-la spécifiquement pour remédier à l'abandon des utilisateurs après l'inscription, mais pas pour corriger une expérience utilisateur (UX) ou des problèmes de performance fondamentalement défaillants.
Installation rapide
Claude Code
Recommandénpx skills add wondelai/skills -a claude-code/plugin add https://github.com/wondelai/skillsgit clone https://github.com/wondelai/skills.git ~/.claude/skills/grow-appCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Grow an App
Your app has users who sign up and then quietly disappear — cohorts decay, daily actives are flat, the activation funnel leaks where it always has, and nothing looks obviously broken. This journey seals the bucket: it turns first-time users into activated, then habitual, then would-miss-it users across eight interactive phases. The agent asks before every decision and records the outcome in docs/, so the work resumes across sessions instead of restarting. Growth here is an engineering and design problem, not a bigger ad budget.
Core Principle
Fix the leaky bucket before pouring in acquisition: habit, activation, and retention come before any growth spend. This skill sequences the eight phases, asks every decision question, and records each choice in docs/. The constituent skills carry the method — invoke them rather than improvising their frameworks.
Journey Map
| Phase | Skill | Question it answers | Artifact |
|---|---|---|---|
| 1 | hooked-ux | Why do users come back without us paying? | Extends docs/PRODUCT.md |
| 2 | improve-retention | Why don't new users reach the loop? | Extends docs/PRODUCT.md |
| 3 | continuous-discovery | What do our own users actually need? | Extends docs/PRODUCT.md + docs/CUSTOMER.md |
| 4 | lean-ux | Which bet is worth building? | Extends docs/EXPERIMENTS.md |
| 5 | inspired-product | Is the team building the right things? | Extends docs/PRODUCT.md |
| 6 | lean-analytics | Which single number tells the truth? | Creates docs/METRICS.md — sets the Rule 8 bar |
| 7 | microinteractions | Does it feel alive in the hand? | Extends docs/DESIGN.md |
| 8 | drive-motivation | Will engagement last, or curdle? | Extends docs/PRODUCT.md |
Phases 1-2 seal the loop and the funnel; 3-5 steer with evidence; 6 is the instrument panel; 7-8 are the finish and the ethical backstop. Take the lean-analytics baseline (Phase 6) early — before the Phase 1-2 fixes land — so every change is read against a pre-change number, then keep updating it. Habit formation is slow: read Phase 1's success against the "5% rule" (a habit has formed when 5%+ of users return unprompted), not a single cohort.
Operating Rules
- Resume first. Before anything else, read
docs/GROW-APP-PLAN.mdand every artifact in the Journey Map. If the tracker exists, summarize the journey state in 3-5 lines and ask which phase to enter. Done when the user has confirmed an entry point. A journey with a tracker is resumed, never restarted. - Intake on first run only. No tracker: run the Intake below, then create
docs/GROW-APP-PLAN.mdwith every phase statusedpending | in-progress | awaiting-evidence | done | deferred: reason | skipped: reason. Done when the tracker exists and the user has confirmed the phase plan. - Phase entry. Announce: what the phase does, the decision it forces, the artifact it produces, rough effort. Offer proceed / skip / defer — phases marked GATE may be deferred, never skipped. Mark the phase
in-progresson proceed. Done when the user chose. - Skill invocation and fallback. Invoke the phase's skill by its slug. If it is not available, offer:
npx skills add wondelai/skills/<slug> --global. If the user declines, run the phase from its Brief — the minimum viable method. State which mode you are in. - In-phase decisions. Ask every question under "Decide with the user" — with concrete options and your recommendation. Record the choice in the tracker's Key Decisions. A decision made silently is a defect.
- Phase exit. Present the draft artifact content for sign-off before writing. On approval: write or extend the docs/ files, update the tracker (status, Key Decisions, Next Actions). Done when the files are written and the phase row shows
done. - Artifact discipline. Read before writing; create a file only if missing, otherwise extend — add or update your sections, preserve everyone else's. Files are UPPERCASE in
docs/. Every recommendation lands as a checkbox or a table row with owner and priority. See references/artifact-templates.md when creating a docs/ file for the first time — create it from the full skeleton (all section headings), then fill the sections your phase names. - Retention before acquisition. Acquisition-oriented phases — the optional cold-start-problem, contagious, and crossing-the-chasm, plus any paid-growth work — stay locked while activation and retention sit below the bar set at intake; unlock them only once the cohort curve clears that bar. Every habit loop and reward must pass the Manipulation Matrix: build only what the maker would use and honestly believes materially improves users. When a tactic needs manufactured anxiety or loss aversion, replace it with one built on real value.
Intake
Run only on first start (no tracker). Ask:
- What does the app do, and what core action does a retained user repeat? (defines the Hook loop and the OMTM)
- What are the current retention numbers — day-1/7/30 or week-4 cohorts? (sets the Rule 8 acquisition bar; feeds lean-analytics)
- Where does the activation funnel leak, and what is the first-run flow? (gates improve-retention)
- Solo/small team or a full product trio (PM, designer, engineer)? (scales continuous-discovery and inspired-product)
- Is the app a network/marketplace product, and do engaged users fail to convert to revenue? (flags optional cold-start-problem / monetizing-innovation)
- What analytics and instrumentation exist today? (gates lean-analytics and every experiment)
- Is retention broken by UX or performance rather than missing engagement? (if yes, route to improve-app first)
Skip heuristics: skip Phase 5 for a solo founder with no team to realign; defer Phase 7 until the loop and activation clear their bars; run Phase 3's cadence degraded if no user access exists yet. Then create the tracker from references/artifact-templates.md with every phase statused, and confirm the plan.
Done when docs/GROW-APP-PLAN.md exists with every phase statused and the user has confirmed the plan.
Phases
Phase 1 — Design the habit loop that brings users back (hooked-ux)
Purpose: Build the engine of return — a Hook loop strong enough that users come back on an internal trigger, not a paid notification.
Brief (fallback): The Hook Model runs Trigger → Action → Variable Reward → Investment. Migrate external triggers (push, email) to internal ones (an emotion — boredom, FOMO, anxiety). Make the action trivially simple. Make the reward variable across tribe/hunt/self. Sequence investment after the reward so it raises switching cost and loads the next trigger. A loop with one weak phase stalls, not half-works.
Invoke: hooked-ux with the core loop and how daily-active users return today. Ask it to (a) map the loop across all four phases, rate each 0-10, and name the weakest, and (b) design honest variable-reward concepts powered by data you already have, each checked against the Manipulation Matrix.
Decide with the user:
- Which internal trigger (emotion) should pull users back — confirm one.
- Which single phase is weakest and gets the highest-leverage fix now — or defer if the loop is already forming (5%+ unprompted return).
- Which reward type to strengthen — tribe (social), hunt (resources), or self (mastery) — rejecting any concept that fails the Manipulation Matrix.
Artifact: Extend docs/PRODUCT.md ## Hook Model (trigger → action → variable reward → investment; weakest phase named). Update the tracker.
Done when: PRODUCT.md names the internal trigger and all four phases, the weakest phase and its fix are recorded, onboarding is re-engineered so a new user completes one full Hook cycle in the first session, and the user picked the fix to ship.
Phase 2 — Fix activation by making the first action almost effortless (improve-retention)
Purpose: Get new users to the loop by making the first meaningful action almost effortless.
Brief (fallback): B=MAP — behavior fires only when Motivation, Ability, and a Prompt converge. Motivation is unreliable; raise Ability instead. Simplicity is capped by the scarcest of six resources: time, money, physical effort, mental effort, social deviance, non-routineness. Shrink the target to a Starter Step that delivers value in under 30s, anchor it to an existing routine, and celebrate the win immediately.
Invoke: improve-retention with the real activation flow step by step and the day-1/7/30 drop-offs. Ask for a B=MAP friction audit rating all six Ability-Chain factors, the scarcest resource named, a Starter Step redesign, and event-based prompt rules.
Decide with the user:
- Which is the scarcest Ability resource for the first action — fix that link first, not the obvious one.
- The Starter Step (tiniest valuable action) and its celebration moment.
- Which time-based prompts convert to event-based, dropping any that fail "would I appreciate this now?".
Artifact: Extend docs/PRODUCT.md ## Activation & Retention Plan (friction/moment | fix | owner | status). Update the tracker.
Done when: the scarcest resource is named, the Starter Step, celebration, and prompt changes are rows with owners, each day-1/7/30 drop-off is mapped to its likely B=MAP failure, and the user approved the fix list.
Phase 3 — Run continuous discovery so you stop guessing (continuous-discovery)
Purpose: Replace generic best-practice with a weekly stream of evidence about your own users.
Brief (fallback): Aim for at least one customer touchpoint per week. Build an Opportunity Solution Tree: outcome at the top → customer opportunities (needs/pains in the customer's words) → candidate solutions/experiments. Never leap outcome→solution. Interviews are story-based ("tell me about the last time you…"), captured as one-page snapshots. Test the riskiest leap-of-faith assumption first, cheaply.
Invoke: continuous-discovery with the retention outcome and known churn patterns. Ask for an Opportunity Solution Tree, a current-state experience map of how churned users try to succeed today, a weekly story-based interview snapshot template, and an assumption map for the next planned feature.
Decide with the user:
- The single outcome at the top of the tree.
- Which two or three opportunities to pursue first.
- The weekly cadence and recruitment mechanism the team can actually sustain — set it now or run degraded.
- The riskiest leap-of-faith assumption inside the next feature (desirability, viability, feasibility, usability) and the cheapest test for it.
Artifact: Extend docs/PRODUCT.md ## Opportunity Solution Tree Notes, ## Outcome Roadmap (outcome/problem | job served | priority | status), and ## Discovery Cadence; extend docs/CUSTOMER.md ## Interview Evidence (date | who | facts | commitment). Update the tracker.
Done when: the tree's outcome and top opportunities are recorded, the cadence is scheduled, the first interview snapshot template exists, and the riskiest assumption has a cheap test designed.
Phase 4 — Replace debate with cheap experiments (lean-ux)
Purpose: Turn opportunities into falsifiable bets settled by behavior, not meetings.
Brief (fallback): Outcomes over outputs — value is the behavior change, not the deliverable. Write a hypothesis: "We believe [outcome] will happen if [persona] achieves [action] with [feature]," with the metric and threshold committed before the test. Match fidelity to risk (a paper prototype with five users finds ~85% of usability issues); reserve A/B tests for tuning a proven concept. When invalidated, remove from the backlog — don't defer.
Invoke: lean-ux with the biggest current design debate or a top discovery opportunity. Ask for three hypothesis statements in the standard format, the lowest-fidelity experiment that could validate the top one, and its pre-committed metric, threshold, and timebox.
Decide with the user:
- Which hypothesis to test first.
- The experiment fidelity — the lowest that answers the actual question.
- The pass/fail line and what leaves the backlog if it fails.
Artifact: Extend docs/EXPERIMENTS.md ## Experiment Cards (hypothesis, type, primary metric + threshold, guardrail, decision rule) and ## Experiment Backlog (idea | ICE | status). Update the tracker.
Done when: at least one experiment card has a pre-committed threshold and decision rule, the backlog is triaged, and the user chose the first test.
Phase 5 — Build the right things with an empowered team (inspired-product)
Purpose: Move the team from feature factory to outcome ownership — problems to solve, not features to ship.
Brief (fallback): Empowered teams get problems, not backlogs, and answer for outcomes. Dual-track: discovery (what's worth building — addressing value, usability, feasibility, viability) runs continuously alongside delivery. Expect 10-20 discovery iterations per shipped feature. Give the team a product vision and an outcome-based roadmap so it can decide autonomously.
Invoke: inspired-product with the top three backlog requests and the current roadmap. Ask for an opportunity assessment of each (objective, target user, problem, success measure, alternatives) and a one-paragraph vision plus a quarter of outcome-based roadmap items.
Decide with the user:
- Which backlog request has the strongest evidence — and which to kill before it reaches a sprint.
- The one-paragraph product vision.
- Whether the roadmap is reframed as problems + key results rather than dated features.
Artifact: Extend docs/PRODUCT.md ## Vision and ## Outcome Roadmap (outcome/problem | job served | priority | status). Update the tracker.
Done when: the vision paragraph exists, each of the three requests has a build/kill verdict, and the roadmap rows are outcomes, not features.
Phase 6 — Measure the one number that actually matters (lean-analytics)
Purpose: Point the whole team at the One Metric That Matters and expose the vanity metrics hiding the decay.
Brief (fallback): A good metric is comparative, a ratio/rate (not a cumulative total), and behavior-changing. Business model dictates which metrics matter; stage dictates sequencing (Empathy → Stickiness → Virality → Revenue → Scale). Weak retention = Stickiness stage, so retention is the OMTM — working a later stage first is the canonical mistake. Draw a line in the sand (target, date, miss response), pair the OMTM with a counter-metric, and cohort the data.
Invoke: lean-analytics with the current dashboard/metrics and the business model. Ask it to flag vanity metrics, pick the Stickiness-stage OMTM plus a counter-metric, design a one-screen dashboard (OMTM big, ≤6 supporting), and a cohorted retention view.
Decide with the user:
- The OMTM and its counter-metric.
- The line in the sand — target, date, pre-committed miss response.
- Which current metrics are retired as vanity.
Artifact: Create docs/METRICS.md with ## Stage & One Metric That Matters, ## KPI Definitions, ## Baselines & Targets, ## Funnel, and ## Cohort Notes. Update the tracker.
Done when: METRICS.md names the OMTM + counter-metric, records the line in the sand with a date, lists cohorted baselines, the vanity metrics are marked retired, the one-screen dashboard is specified, and the retention bar for the Rule 8 acquisition gate is set.
Phase 7 — Polish the micro-moments that make it feel alive (microinteractions)
Purpose: Close the gap between an app people tolerate and one they love, in the moments they touch daily.
Brief (fallback): Every microinteraction has Trigger → Rules → Feedback → Loops & Modes. Feedback is immediate (<100ms for direct manipulation) and proportionate; animate the element the user touched over a separate toast. Map every state: empty, loading, partial, error, disabled, double-tap. Invest in one or two signature moments that pass the removal test; use long loops to retire hints for power users.
Invoke: microinteractions with the five most-used interactions. Ask for a Trigger/Rules/Feedback/Loops audit of each, the sub-100ms feedback and missing edge-case states, and one signature moment implemented in real code.
Decide with the user:
- Which five interactions to audit.
- Which one becomes the signature moment (removal test applied).
- Which edge-case states to implement first.
Artifact: Extend docs/DESIGN.md ## Microinteraction Inventory (interaction | trigger/rules/feedback/loops | fix | status). Update the tracker.
Done when: the five interactions are in the inventory with their missing states and fixes, the signature moment is chosen, and each fix has a status.
Phase 8 — Sustain engagement with intrinsic motivation (drive-motivation)
Purpose: Keep the loops from curdling — engagement that runs on Autonomy, Mastery, and Purpose instead of exploitation.
Brief (fallback): For any task needing cognitive effort, "if-then" rewards crush intrinsic motivation. Lasting engagement is Autonomy (choice over what/when/how/with whom), Mastery (visible progress, flow-calibrated challenge), and Purpose (why it matters). Autonomy killers: forced tutorials, unskippable steps, mandatory notifications. Reserve rewards for meaningful milestones; prefer "now-that" recognition over "if-then" bargains.
Invoke: drive-motivation with the app's gamification, streaks, points, and notification patterns. Ask for an AMP audit rated 0-10, every autonomy violation flagged, the point at which streaks tipped into loss aversion, and a progression redesign around real mastery and purpose.
Decide with the user:
- Which autonomy violations to remove (forced/unskippable steps).
- Which "if-then" rewards convert to "now-that" recognition.
- Whether any streak/points mechanic exploits loss aversion and must change.
Artifact: Extend docs/PRODUCT.md ## Activation & Retention Plan with AMP-audit rows (violation/finding | fix | owner | status). Update the tracker.
Done when: the AMP score and every autonomy violation are recorded, the reward/streak fixes are rows with owners, and the loops pass the Manipulation Matrix from Rule 8.
Optional Phases
| Skill | Add when | Artifact |
|---|---|---|
| cold-start-problem | the app is a network or marketplace product | Extends docs/PRODUCT.md |
| monetizing-innovation | engaged users do not translate into revenue | Extends docs/OFFER.md |
| contagious | users love the app but never mention it | Extends docs/MARKETING.md |
| crossing-the-chasm | growth stalls at the early-adopter boundary | Extends docs/STRATEGY.md |
| jobs-to-be-done | usage patterns say the app is hired for a different job | Extends docs/CUSTOMER.md |
Optional phases follow the same operating rules; insert where the Add-when condition first becomes true. The acquisition-leaning ones — cold-start-problem, contagious, crossing-the-chasm — stay locked behind Rule 8 until retention clears the bar.
Common Mistakes
| Mistake | Fix |
|---|---|
| Buying growth before fixing retention | Pass the Stickiness gate (a flattening cohort curve) before any acquisition spend; keep acquisition phases locked per Rule 8. |
| Relying on external triggers forever | Migrate to an internal trigger via hooked-ux; treat notifications as scaffolding, not the load-bearing wall. |
| Optimizing the wrong Ability-Chain link | Rate all six factors in improve-retention and fix the scarcest resource, not the most obvious one. |
| Jumping from outcome straight to solution | Build the Opportunity Solution Tree in continuous-discovery first; the obvious feature is often the worst of five. |
| Measuring outputs, not outcomes | Instrument every release; in lean-ux and inspired-product, success is a change in user behavior, not stories shipped. |
| Gamifying with points for everything | Reserve rewards for meaningful milestones and run the drive-motivation AMP audit; "if-then" rewards crowd out your power users. |
Completing the Journey
- PRODUCT.md holds a Hook loop with the weakest phase fixed, an activation Starter Step, a living Opportunity Solution Tree, and an AMP-clean engagement design.
- METRICS.md names the Stickiness-stage OMTM plus a counter-metric with a line in the sand (target, date, miss response).
- At least one lean-ux experiment has resolved with a recorded verdict, and invalidated ideas are out of the backlog.
- The retention bar set at intake is met — or the remaining gap is quantified — before any acquisition phase runs.
Close the tracker: every phase done or skipped, with Next Actions carried into PRODUCT.md, METRICS.md, and EXPERIMENTS.md. Then route forward:
- When engagement mechanics cannot fix a product held back by broken UX or performance, continue with
improve-app. - When the app is sticking and the business around it must keep pace — revenue, channels, operations — continue with
grow-business.
Dépôt GitHub
Questions fréquentes
Qu’est-ce que le Skill grow-app ?
grow-app est un Skill Claude créé par wondelai. Un Skill regroupe des instructions et des ressources que Claude charge à la demande pour effectuer des tâches liées à grow-app sans consigne supplémentaire.
Comment installer grow-app ?
Utilisez les commandes d’installation de cette page : ajoutez grow-app à Claude Code comme plugin ou clonez son dépôt dans votre dossier skills, puis redémarrez Claude pour charger le Skill.
À quelle catégorie appartient grow-app ?
grow-app appartient à la catégorie Méta.
grow-app est-il gratuit ?
Oui. grow-app est référencé sur AIMCP et son installation est gratuite.
Compétences associées
Cette compétence propose une configuration éprouvée en production pour Content Collections, un outil axé sur TypeScript qui transforme des fichiers Markdown/MDX en collections de données typées de manière sûre avec une validation Zod. Utilisez-la lors de la création de blogs, de sites de documentation ou d'applications Vite + React riches en contenu pour garantir la sécurité de typage et la validation automatique du contenu. Elle couvre tout, de la configuration du plugin Vite et de la compilation MDX à l'optimisation des déploiements et la validation des schémas.
Cette compétence permet aux développeurs de créer des applications avec la plateforme de marchés prédictifs Polymarket, incluant l'intégration d'API pour le trading et les données de marché. Elle fournit également une diffusion de données en temps réel via WebSocket pour surveiller les transactions en direct et l'activité du marché. Utilisez-la pour mettre en œuvre des stratégies de trading ou pour créer des outils traitant les mises à jour de marché en direct.
Cette compétence aide les développeurs à créer des plugins OpenCode qui s'interconnectent avec plus de 25 types d'événements tels que les commandes, les fichiers et les opérations LSP. Elle fournit la structure du plugin, les spécifications de l'API événementielle et les modèles d'implémentation pour les modules JavaScript/TypeScript. Utilisez-la lorsque vous avez besoin d'intercepter, de surveiller ou d'étendre le cycle de vie de l'assistant IA OpenCode avec une logique personnalisée pilotée par les événements.
SGLang est un framework de service LLM haute performance spécialisé dans la génération rapide et structurée pour les workflows JSON, regex et agentiques grâce à son cache de préfixe RadixAttention. Il offre une inférence nettement plus rapide, particulièrement pour les tâches avec des préfixes répétés, ce qui le rend idéal pour les sorties complexes et structurées ainsi que les conversations multi-tours. Choisissez SGLang plutôt que des alternatives comme vLLM lorsque vous avez besoin d'un décodage contraint ou que vous construisez des applications avec un partage étendu de préfixes.
