---
name: problems-first
description: Pressure-test feature requests, product ideas, PRDs, and user stories in the Problem Space before writing specs or code. Peel back surface solution requests with the 5 Whys, audit user workarounds and situational context, filter scope bloat on the Kano curve, distinguish local-maximum tweaks from higher-ceiling workflows, and issue a Stop, Prototype-to-Learn, or Build-to-Earn gate verdict. Use whenever a user asks to scope a feature, write a PRD or user story, validate a startup or product bet, prioritize a backlog, plan an MVP, or jump straight into building a new app or workflow.
metadata:
  author: Andrew Durden
  version: 1.1.0
  visibility: public
  website: https://andrewdurden.com/skills/problems-first/SKILL.md
license: MIT
---
# Problems First

Coding agents have dropped the cost of generating first-draft interfaces and backend code near zero, which makes building and scaling the wrong thing fast. Run this pre-code discovery workflow to translate surface feature requests into real human friction, check whether you are polishing a low-ceiling flow or jumping to a better workflow, cut indifferent scope, and define the fastest falsifiable test before committing engineering capacity.

## When to Use (and When to Shorten or Skip)

- **Use in full:** Scoping stakeholder feature requests ("add an AI assistant", "send a daily digest"), critiquing PRDs or MVP plans, or deciding whether an idea deserves multi-week engineering investment.
- **Shorten to 2 minutes:** Small usability fixes, broken states, or obvious bugs where user friction is already observed in support tickets or session replays; state the broken moment and ship the fix.
- **Hospitality, Delight & Latent-Need Exceptions:** Not every worthwhile build starts from an ugly spreadsheet workaround. People rarely maintain a manual workaround when they have given up on an inaccessible task, and personal hospitality or creative craft (such as a custom event camera, a calm living-room display, or a neighborhood field guide) earns its keep through warmth, taste, and shared joy rather than utilitarian pain relief. In those cases, replace the workaround audit with a **consideration check**: confirm the delight stays out of the way of the user's primary task.

## Discovery Workflow

Track progress through these five gates before writing production specs or code:

- [ ] Step 1: Peel back 3 to 5 Whys from Solution Space into Problem Space
- [ ] Step 2: Anchor situational context, frequency, and how people cope today
- [ ] Step 3: Check the ceiling (incremental tweak vs. workflow redesign)
- [ ] Step 4: Classify candidate scope on the Kano curve and cut the indifferent middle
- [ ] Step 5: Issue one explicit gate verdict (`Stop / Simplify`, `Prototype to Learn`, or `Build to Earn`)

### Step 1: Cross From Solution Space Into Problem Space

Users and stakeholders naturally ask for solutions wrapped in UI nouns ("build a dashboard", "add a chatbot", "give me a notification toggle"). Treat the initial noun as a clue and ask **Why?** until you hit a concrete human or commercial consequence (typically three to five layers, stopping earlier if the root friction is already plain):

- **Surface request (Solution Space):** "Add an AI bot that summarizes every customer call into Slack."
- **Why 1?** "Because product owners rarely read 45-minute call transcripts."
- **Why 2?** "Because spotting repeated onboarding blockers takes three hours every Friday."
- **Why 3?** "Because onboarding regressions slip through two sprints unnoticed and spike week-one churn."
- **Root problem (Problem Space):** Catching recurring week-one onboarding blockers before sprint planning locks, without hours of manual transcript skimming.

Rewrite the problem statement without naming any UI widget, database table, or AI model.

### Step 2: Anchor Situational Context and Coping Behavior

Audit four concrete signals around the moment of use:

1. **Physical and cognitive environment:** Identify where the person is standing or sitting, what device they hold, how noisy or time-pressured the moment is, and whether connectivity or attention is fragmented.
2. **Current coping behavior (or abandonment):** Look for manual spreadsheets, copy-pasted notes, phone photos, or five open browser tabs. If no workaround exists, check whether users abandoned the goal out of frustration, lacked access, or are seeking hospitality/creative delight rather than pain relief.
3. **Importance vs. current satisfaction:** Prioritize moments where current options actively fail, exclude, or frustrate people; skip low-friction steps that already work well enough.
4. **Audience surface check:** Keep one calm primary action on the surface for first-time or casual users while tucking power-user controls behind progressive disclosure.

### Step 3: Check the Ceiling Before Polishing

Incremental optimization only improves the flow you already have. Before scoping tweaks to an existing screen, challenge the approach:

1. **Name the current flow's ceiling:** Ask what happens if the requested tweak works as imagined. Does it unlock a meaningful outcome shift, or only shave seconds off a broken step?
2. **Test a blank-sheet question:** Ask how a team starting fresh today would remove the user's manual work altogether instead of adding another filter, toggle, or alert.
3. **Explore cheaply when uncertain:** When a broader redesign is plausible but unproven, compare two lightweight workflow prototypes before hardening production architecture.

### Step 4: Cut Indifferent Scope on the Kano Curve

Classify proposed capabilities and cut the indifferent middle so fast code generation does not turn into cognitive clutter and maintenance debt:

| Kano Tier | What It Means for the User | Scoping Rule |
| :--- | :--- | :--- |
| **Must-Haves** | Baseline trust, accuracy, privacy, and reliability; breaking one causes abandonment | Keep non-negotiable and verify with deterministic tests rather than unguarded model outputs |
| **Performance** | The 1 to 2 speed, clarity, or accuracy axes users actively compare | Pick the primary axis that matters most in Step 2's environment and win it cleanly |
| **Delighters** | Thoughtful care or workflow compression users did not think to request | Invest saved build time in one genuine Delighter that never blocks the core path |
| **Indifferent Bloat** | Speculative settings, edge-case toggles, and "nice-to-have" checkboxes | Cut early; every unvalidated feature taxes comprehension and future iteration speed |

### Step 5: Issue One Gate Verdict and Falsification Signal

Conclude with one decision, what evidence is still missing, and a concrete test:

- **`Stop / Simplify`:** A clearer label, smarter default sort, or policy tweak solves most of the friction with zero new surface area.
- **`Prototype to Learn`:** High upside with real value or comprehension uncertainty. Build a 1-to-3-day interactive prototype with realistic data, watch 5 target users attempt the task, and observe where they hesitate.
- **`Build to Earn`:** User pull and comprehension are already supported by evidence; invest in production architecture, security boundaries, caching, and regression gates.

### Worked Example Where This Skill Leads to Doing Less (`Stop / Simplify`)

- **Stakeholder Ask:** "Build a custom saved-search alert center so first-time shoppers get emailed whenever listings change."
- **Peeled Root Problem:** First-time visitors bounce on page one because the default `Price: Low to High` sort buries the listings most people want beneath cheap ones few people choose, making the catalog look confusing.
- **Verdict (`Stop / Simplify`):** Skip the alert center for anonymous first-time visitors (who rarely register before finding something they like) and test a smarter geographic + session-intent default sort on visit one instead.

## Output Template

```markdown
# Problem & Scoping Brief: [Short Initiative Name]

## 1. Root Problem & Context
- **Original Ask:** [Verbatim feature or stakeholder request]
- **Peeled Root Problem:** [UI-free statement of the human job, stakes, and moment]
- **Context & Coping Behavior:** [Device/time constraints + current workaround, abandonment, or delight goal]
- **Evidence Still Missing:** [What this brief cannot yet conclude without user observation]

## 2. Ceiling Check
- **Incremental Option:** [Local tweak and its realistic ceiling]
- **Workflow Redesign Option:** [Alternative that removes manual steps altogether]

## 3. Kano Scope Filter
- **Must-Haves:** [1 to 3 non-negotiable trust/reliability invariants]
- **Primary Performance Axis:** [Single speed/accuracy/clarity metric to prioritize]
- **Earned Delighter (Optional):** [One considerate detail that stays out of the way]
- **Explicitly Cut:** [Features or settings excluded from this cycle]

## 4. Gate Verdict & Falsification Test
- **Verdict:** [Stop / Simplify / Prototype to Learn / Build to Earn]
- **Fastest Test:** [Smallest change or prototype to put in front of users]
- **Falsification Signal:** [Observable behavior that would change our mind]
```

## Verification Check

1. **Zero Solution Nouns in Root Problem:** Confirm the root problem statement contains no UI widgets, model names, or database terms.
2. **Honest Evidence Boundary:** Confirm `Evidence Still Missing` names what remains unverified rather than presenting hypotheses as facts.
3. **Observable Falsification Signal:** Confirm the test watches real user behavior rather than asking "would you use this?"

## Gotchas

- **Mistaking polite compliments for pull:** Watch whether people complete the task or revert to their old habit rather than relying on interview praise.
- **Promoting discovery spikes straight to production:** Prototype code optimizes for learning speed; add auth isolation, validation, and tests before shipping to real accounts.
