# Retention Loop Designer > Designs the habit-forming mechanism that brings users back without prompting, using the Hook Canvas framework. Spawns 4 parallel agents to design each phase of the hook simultaneously — Trigger, Action, Variable Reward, Investment. Use when users aren't returning on their own, when retention is flat, or when you want to build compounding engagement into the product. Produces retention-loop.md wit… - **Type:** Skill - **Install:** `agentstack add skill-gvkhosla-founder-skills-retention-loop-designer` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [gvkhosla](https://agentstack.voostack.com/s/gvkhosla) - **Installs:** 0 - **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [gvkhosla](https://github.com/gvkhosla) - **Source:** https://github.com/gvkhosla/founder-skills/tree/main/legacy/skills/pmf/retention-loop-designer - **Website:** https://fskills.xyz ## Install ```sh agentstack add skill-gvkhosla-founder-skills-retention-loop-designer ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Retention Loop Designer ## The Core Insight Most products rely on external triggers to bring users back: push notifications, email reminders, re-engagement campaigns. These work — until users turn them off or start ignoring them. The best products create internal triggers: the user feels something — boredom, anxiety, curiosity, social pressure — and the product comes to mind as the solution. No notification needed. This is habit. The difference between a product people return to willingly and one they return to because they were nudged is the difference between retention that scales and retention that requires constant marketing spend to maintain. ## Quick Start ## Human-First Response Default: answer in chat first with **Verdict**, **Do this now**, and **Details saved**. Keep chat under 150 words unless asked; do not paste the full artifact. Write/update the requested `.md` file as the durable record for agents. Say: **"Design my retention loop"** or **"Why aren't users coming back?"** or **"Build my hook"** Output: `retention-loop.md` — a complete Hook Canvas for your product with specific implementation priorities. --- ## The Hook Canvas Nir Eyal's framework from *Hooked*. Four phases that, when working together, create habits. ``` ┌─────────────────────────────────────┐ │ HOOK │ │ │ │ TRIGGER → ACTION → REWARD → INVEST │ │ ↑ │ │ │ └────────────────────────┘ │ │ (investment feeds next │ │ trigger — the loop) │ └─────────────────────────────────────┘ ``` --- ## Parallel Execution Each phase of the Hook Canvas is independent — all four can be designed simultaneously. **Before spawning, gather:** - `founder-context.md` — customer profile, product description, current stage - `customer-profile.md` — the painful moment, what they fear, their existing habits - `user-flows.md` — the core user journey (if it exists) - Any cycle notes about why users do or don't return **Spawn these 4 agents simultaneously:** **Agent 1 — Trigger Designer** Context: Customer profile + product description + current usage patterns Task: Design the trigger system for this product. Trigger types: - **External triggers:** Push notification, email, SMS, badge — someone tells the user to come back - **Internal triggers:** An emotion or state that cues the behavior (boredom → Instagram, anxiety → checking email, FOMO → Slack) The goal: design external triggers that create internal ones over time. Questions to answer: - What emotional state does the target user regularly experience that this product could solve? - What existing daily habit could this product be attached to? (habit stacking) - What external trigger is low-friction, high-relevance, and not annoying? - How many uses does it take before the internal trigger starts forming? Returns: Primary internal trigger (the emotion) + external trigger design (what, when, frequency) + the habit stacking opportunity (what existing behavior to attach to) **Agent 2 — Action Designer** Context: Customer profile + product's core value + user flow Task: Design the simplest possible action that initiates the habit loop. Fogg Behavior Model: Behavior = Motivation × Ability × Prompt The action must be: - Simple enough to do in the triggered state (low ability required) - Directly connected to the reward - The minimum action that moves toward value Questions to answer: - What is the single simplest action a user takes to start getting value? - What friction currently exists between trigger and action? (Each friction point is a drop-off) - Can the action be shortened to under 30 seconds? - Is there a "minimum viable engagement" version of the core action? Returns: The designed action (specific, described step-by-step) + current friction map + specific friction to remove first **Agent 3 — Variable Reward Designer** Context: Product description + customer profile + what users currently feel after using it Task: Design the variable reward system. Variable rewards are the engine of habits. Predictable rewards lose their pull quickly — variability creates the compulsion to return. Three reward types: - **Tribe rewards:** Social validation, recognition, belonging (likes, responses, leaderboard position) - **Hunt rewards:** The search for information, resources, or outcomes (feed refresh, search results, new content) - **Self rewards:** Mastery, completion, personal achievement (streak, level up, goal met) Questions to answer: - Which reward type fits this product's core value? (social products → tribe; productivity → self; content → hunt) - Where is the moment of variability in the current product? (If there isn't one, design it) - How can the reward be made variable without feeling random or cheap? - What "near miss" mechanism creates the desire to try again? Returns: Primary reward type + specific reward design + variability mechanism + the moment of delight to design toward **Agent 4 — Investment Designer** Context: Product description + customer profile + current data/content users put into the product Task: Design the investment phase that makes the product more valuable with use and feeds the next trigger. Investments are what users put INTO the product: data, content, relationships, customization, reputation, skill. Investments: 1. Make the product better for the user over time (switching cost increases) 2. Load the next trigger (the investment creates the context for the next return) Questions to answer: - What can users store, create, or build in this product that makes it more valuable to them over time? - Does the product currently get smarter/better the more a user uses it? If not, how could it? - What investment creates a "loaded next trigger"? (e.g., posting content → notification when someone responds) - What would make a user feel that leaving means losing something valuable? Returns: Primary investment type + how it loads the next trigger + specific feature that enables the investment + switching cost it creates **Wait for all 4 agents. Orchestrator assembles the complete Hook.** --- ## Synthesis (Orchestrator Only) Assemble the four agent returns into a complete Hook Canvas. Check that the loop is coherent: - Does the investment feed the next trigger? (The loop must close) - Is the action simple enough to do in the triggered state? - Is the reward genuinely variable, or just reliably pleasant? - Is the investment something users would actually do (not just something clever in theory)? **Write `retention-loop.md`:** ```markdown # Retention Loop — [Product Name] — [YYYY-MM-DD] ## The Hook **Internal Trigger:** [The emotion/state that cues return — e.g., "anxiety about forgetting something"] **External Trigger:** [What, when, and how — e.g., "weekly digest email every Monday 9am showing last week's stats"] **Action:** [The simplest behavior — described step by step] **Friction to remove first:** [The single highest-friction point between trigger and action] **Variable Reward:** [Type + specific design — e.g., "Hunt reward: which waste category changed most this week?"] **The variability:** [What changes each return that makes it worth coming back] **Investment:** [What the user puts in — e.g., "daily waste logs that build a personal historical record"] **How it loads the next trigger:** [The loop closure — e.g., "logs accumulate → weekly summary → curiosity about trend → return"] ## Implementation Priorities The hook only works if all 4 phases work. Build in this order: 1. **[Highest-friction fix]** — Remove the biggest obstacle between trigger and action (1 day) 2. **[Investment feature]** — Build the accumulation mechanism that makes the loop self-sustaining (1 week) 3. **[Variable reward]** — Introduce the variability that makes return worth it (3 days) 4. **[External trigger]** — Set up the notification/email trigger (1 day — do last, after loop works) ## What to Watch - D7 retention: should improve within 2 cycles of implementing the investment phase - Time-between-sessions: should decrease as internal trigger strengthens - Session-initiating action: track what % of sessions start from external vs. internal trigger over time ``` --- ## Sequential Fallback (Codex) Design each Hook phase in sequence: 1. Trigger design 2. Action design 3. Variable reward design 4. Investment design 5. Assemble and write `retention-loop.md` --- ## Related Skills - Use **pmf-signal-reader** before this — confirms retention is the signal to amplify - Use **north-star-definer** to align the retention metric with the north star - Use **ux-flow-designer** (Design phase) alongside — the hook lives inside the UX flow - Use **build-cycle** — retention improvement tracked across subsequent cycles ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [gvkhosla](https://github.com/gvkhosla) - **Source:** [gvkhosla/founder-skills](https://github.com/gvkhosla/founder-skills) - **License:** MIT - **Homepage:** https://fskills.xyz Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-gvkhosla-founder-skills-retention-loop-designer - Seller: https://agentstack.voostack.com/s/gvkhosla - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.