improve-retention
Diagnose and fix retention problems using behavior design (B=MAP)
Install / Use
npx skills add wondelai/skills --skill improve-retentionInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Human ResourcesSupported Platforms
Our assessment of improve-retention
improve-retention scores 90/100 on our quality scale, 20th of 46 Human Resources skills we index (top 44%).
Its SKILL.md is 15 KB long, well organised into 24 sections with 2 code examples: a thorough specification that gives an agent plenty to work with.
With 2,248 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so improve-retention is actively maintained.
- Our last check on 2026-10-30 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-10-01. Automated pattern scan on 2026-09-30. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
improve-retention compared with similar skills
All 4 of these similar skills score higher than improve-retention; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| improve-retention (this skill)by wondelai | 90 | 2.2k | 2mo ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| Claude-Code-Game-Studiosby Donchitos | 100 | 39.7k | 7d ago | CLAUDE.md |
| x402by browser-use | 100 | 121.0k | 3d ago | SKILL.md |
| algorithmic-artby anthropics | 100 | 177.9k | 55d ago | SKILL.md |
Frequently asked questions
- How do I install improve-retention?
- Run
npx skills add wondelai/skills --skill improve-retention. The install tabs above show the steps for each supported agent. - Which AI agents does improve-retention work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is improve-retention safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is improve-retention still maintained?
- The repository was last updated about 2 months ago, so improve-retention is actively maintained.
Skill content
View source on GitHubname: improve-retention description: 'Diagnose and fix retention problems using behavior design (B=MAP). Use when the user mentions "users sign up but dont stick around", "activation rate", "onboarding friction", "retention metrics", "why users dont complete", "churn analysis", or "aha moment". Also trigger when analyzing cohort retention curves, designing activation milestones, reducing time-to-value for new users, or investigating why users quit after their first session. Covers the Ability Chain, prompt design, and tiny behaviors that compound. For habit loops and variable rewards, see hooked-ux. For intrinsic motivation, see drive-motivation.' license: MIT metadata: author: wondelai version: "1.4.0"
Behavior Design Framework
Framework for designing products that reliably change behavior. Behavior is not about willpower or motivation — it is a design problem with a predictable equation.
Core Principle
The Fogg Behavior Model = B=MAP. Behavior happens when Motivation, Ability, and a Prompt converge at the same moment.
HIGH ┃
┃ ★ Behavior happens
┃ (above the Action Line)
┃
Motivation ┃━━━━━━━━━━━━━━━━━━━━━━━ ← Action Line
┃
┃ ✗ Behavior fails
┃ (below the Action Line)
LOW ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━
HARD EASY
Ability
The Action Line: When motivation and ability are sufficient, a prompt causes the behavior; below the line, no prompt works. High motivation compensates for low ability and vice versa. The reliable strategy is making behaviors easier (move right), not pumping up motivation (move up).
See: references/behavior-model.md when you need the curve mechanics behind this model — the full Action Line math, behavior types (dot/span/path), and a step-by-step failure diagnostic for a behavior that isn't happening.
Scoring
Goal: 10/10. The six Quick Diagnostic rows are the single source of score-movers. Rate each pass/fail, then start at 10 and subtract per failing row: low motivation or below the Action Line (rows 1-2) cost -2 each; prompts, celebration, bottleneck, and scaling (rows 3-6) cost -1.5 each. A design that passes all six scores 10; one that fails every row scores 0. Map to bands: 9-10 = behavior reliably crosses the Action Line at low motivation, prompts are event/anchor-tied, key actions are celebrated; 5-6 = depends on a motivation spike or optimizes a non-bottleneck factor; <=3 = core action below the Action Line, prompts are spam, no habit wiring. Always state the score and name the specific failing rows.
The Three Elements
1. Motivation
Core concept: Motivation is the energy for action, driven by three core motivators, each with two sides: Sensation (pleasure/pain), Anticipation (hope/fear), Belonging (acceptance/rejection). It is powerful but unreliable.
Why it works: Motivation comes in waves — it spikes (New Year's resolutions, product launches) and crashes (day 3, week 2). Products that depend on high motivation fail when the wave recedes; the best designs work at the trough.
Key insights:
- "Motivation is unreliable. Ability is not." — BJ Fogg
- Design for low-motivation moments, not peak excitement
- Motivation-first tactics (inspiring videos, aspirational messaging) produce spikes, not sustained behavior
- Match required motivation to behavior difficulty — hard behaviors need high motivation
Product applications:
| Context | Application | Example | |---------|-------------|---------| | Onboarding | Don't count on the new-user spike lasting | First actions work even when excitement fades | | Re-engagement | Assume returning users have low motivation | Show immediate value before asking for effort | | Messaging | Tap the right motivator | Social fitness → belonging; financial tool → hope |
Copy patterns:
- "Takes 30 seconds" (signals ease, lowers motivation needed)
- "Join 50,000 teams who..." (belonging motivator)
- "Don't lose your 7-day streak" (anticipation/fear motivator)
Ethical boundary: A fear motivator (the streak pattern above) is fair only when the loss is real and user-owned (their data, their progress); never invent a loss that exists solely to drive a session.
See: references/motivation-waves.md for the three motivators, motivation waves, and designing for troughs.
2. Ability
Core concept: Ability is the capacity to do the behavior — a function of the scarcest resource across six factors (the Ability Chain). If any single link is too weak, the behavior breaks.
Why it works: Unlike motivation, ability can be systematically engineered: every removed field, eliminated step, and preset default moves the behavior right on the model, crossing the Action Line even at low motivation. The Ability Chain gives you the diagnostic — find the weakest link and fix it.
Key insights:
- Six factors: Time, Money, Physical Effort, Mental Effort, Social Deviance, Non-Routine
- Simplicity is a function of the scarcest resource — find the bottleneck, not the most obvious factor
- "Simplicity changes behavior" — BJ Fogg
- Starter Steps: shrink the behavior to the tiniest version (2 minutes → 30 seconds → one field)
- Defaults are the most powerful ability tool — users rarely change them
Product applications:
| Context | Application | Example | |---------|-------------|---------| | Signup | Cut cost across all six factors | One-click SSO removes time, mental effort, non-routine | | Core action | Fix the weakest link | Mental-effort bottleneck → smart defaults and templates | | Enterprise adoption | Address social deviance | "Your team already uses this" reduces social risk |
Copy patterns:
- "One click to get started" (time + physical effort)
- "No technical skills needed" (mental effort)
- "Works just like tools you already use" (non-routine)
Ethical boundary: Reduce friction only on genuinely valuable behaviors — never make it too easy to overspend, over-share, or delete important data without confirmation.
See: references/ability-chain.md for the six factors in detail, friction audit templates, and simplification strategies.
3. Prompt
Core concept: The prompt says "do it now." Without one, behavior doesn't happen regardless of motivation and ability. Three types: Person Prompts (internal reminders), Context Prompts (environmental cues), Action Prompts (designed triggers from the product).
Why it works: Teams assume motivation + ability is enough — it isn't, not without a well-timed prompt. But prompts only work above the Action Line: a push notification to someone lacking motivation or ability is spam.
Key insights:
- A prompt at the wrong moment is noise; at the right moment, magic
- Anchor moments tie new behaviors to existing routines ("After I open Slack, I will...")
- Prompt fatigue is real — every unnecessary prompt degrades the value of future ones
Product applications:
| Context | Application | Example | |---------|-------------|---------| | Notifications | Prompt only above the Action Line | Send digest when there's content to review, not on a schedule | | Re-engagement | Tie prompts to real events | "Your report is ready" (event-based, not time-based) | | Feature discovery | Prompt when motivation and ability align | Feature tour appears when user hits the problem it solves |
Copy patterns:
- "Your weekly report is ready" (context prompt — real event)
- "One thing left to complete your setup" (action prompt with progress)
- Never: "We miss you!" (product need, not user need)
Ethical boundary: Every prompt must pass the test "Would I appreciate receiving this right now?" — if it serves a product metric (DAU, re-engagement) but not the user's current goal, cut it.
See: references/prompt-design.md for prompt types, timing strategies, notification design, and anchor moments.
Tiny Habits Method
The practical application of B=MAP: make behaviors so small they need almost no motivation, anchor them to existing routines, and celebrate immediately.
The Recipe
After I [ANCHOR MOMENT], I will [TINY BEHAVIOR], then I [CELEBRATION].
- Anchor Moment: an existing routine that reliably happens (opening an app, finishing a meeting, morning coffee).
- Tiny Behavior: the smallest version of the target behavior — not "write a report" but "open the report template."
- Celebration: an immediate positive emotion that wires the habit. Repetition alone isn't enough — you need the feeling of success.
Starter Steps
Every target behavior has a Starter Step — the tiniest meaningful version:
| Target Behavior | Starter Step | Why It Works | |----------------|--------------|--------------| | Complete onboarding | Fill in one field | Momentum from completion | | Use analytics daily | Open the dashboard | Seeing data creates curiosity | | Collaborate with team | Send one comment | Social reciprocity kicks in |
Scaling Behaviors
Once wired, tiny behaviors grow naturally: open dashboard → check a few metrics → customize → automatic morning habit. Never force scaling — let motivation and momentum drive expansion. The tiny version is the foundation, not a failure.
See: references/tiny-habits.md for the full Tiny Habits recipe, celebration techniques, and scaling patterns.
Behavior Design Process
Fogg's systematic process for lasting behavior change:
Step 1: Clarify the Aspiration
What outcome does the user want — their aspiration, not the product's goal ("stay on top of my team's progress", not "increase DAU").
Step 2: Explore Behavior Options
List all possible behaviors that could achieve the aspiration. Be exhaustive — don't commit yet.
Step 3: Match Behaviors
Assess each for motivation and ease; plot on a 2×2 of impact vs. feasibility (Focus Mapping).
Step 4: Start Tiny
Shrink the best-matched behavior to its Starter Step; design the prompt; add celebration.
Step 5: Optimize
Expand once wired. Fix bottlenecks with the Ability Chain; refine prompt timing from data.
See: references/product-applications.md when applying this process to a specific category — B=MAP mapped to SaaS onboarding, mobile, e-commerce, health, and education with per-category motivation timelines and bottlenecks.
The Action Line
Moving Behaviors Above the Action Line
- Increase Ability (move right) — remove steps, pre-fill, defaults, templates, wizards. The most reliable approach.
- Find better Prompts — anchor to existing routines; event-based beats time-based; trigger when motivation is naturally higher.
- Increasing Motivation (move up) is unreliable — if you need motivation tactics, the behavior is probably too hard.
Retention Diagnostics with B=MAP
Map B=MAP to product metrics:
| Metric | B=MAP Diagnosis | Action | |--------|----------------|--------| | Low activation | First action below the Action Line | Shrink onboarding to Starter Step; fix weakest Ability Chain link | | Day-1 drop-off | Prompt failed or mistimed | Redesign first-day prompts; anchor to an existing routine | | Day-7 drop-off | Motivation wave receded, behavior too hard | Reduce core action difficulty | | Day-30 drop-off | Habit didn't form, no internal prompt | Create tiny habit recipe; add celebration loops | | Low feature adoption | Feature below the Action Line for most users | Friction-audit it; prompt only when motivation is present | | Notification fatigue | Prompts sent below the Action Line | Cut volume; send only with motivation + ability |
See: [references/case-studies.md](ref
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
Claude-Code-Game-Studios
39.7kTurn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
x402
121.0kSet up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key.
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
