<objective> Analyze artifacts against the One-Click Promise philosophy to identify unnecessary complexity, missing defaults, excessive user decisions, and opportunities for automation. Output specific violations with actionable improvements. </objective> <scope> **What to review:** - Product Requirements Documents (PRDs) - User flows and interaction designs - Prompts and AI agent behavior specs - Onboarding sequences - Form designs and data collection flows - Error handling and recovery flows

What this skill does NOT cover:

•Code quality or architecture review
•Visual design aesthetics
•Technical implementation details </scope>

<process> 1. **Read the reference** - Load `references/principles.md` to have the full framework

•
Identify the artifact type - PRD, user flow, prompt, agent spec, or other
•
Apply the Decision Test first - For each user action required:

"Would a thoughtful person regret it if we just did this for them?"
- •No → Flag as automation opportunity
- •Yes → Acceptable decision point, but check if it can be a confirmation vs. open question
•
Scan for principle violations - Check each of the 10 principles against the artifact
•
For AI/agent artifacts - Also apply the 7 Agent Guidelines
•
Output the review using this structure:

code

## One-Click Review

**Artifact:** [name/description]
**Type:** [PRD | User Flow | Prompt | Agent Spec | Other]

### Decision Test Results
[List each user action and whether it passes the test]

### Violations Found
[For each violation:]
- **Principle X:** [principle name]
  - **Where:** [specific location in artifact]
  - **Issue:** [what's wrong]
  - **Fix:** [specific improvement]

### Automation Opportunities
[Things that could be done for the user without asking]

### Quick Wins
[Easiest changes with highest impact]

### Summary
[One paragraph: overall assessment and priority recommendations]

</process>

<evaluation_lens> When reviewing, ask these questions:

•Where is the user being asked something we could figure out ourselves?
•What fields have no default when one could be inferred?
•Where does jargon appear instead of plain language?
•What multi-step processes could be single actions?
•Where do errors require user problem-solving vs. auto-resolution?
•What decisions are being asked that the user doesn't care about?
•Is confirmation used instead of interrogation where possible? </evaluation_lens>

<success_criteria> A complete review:

•Applies the Decision Test to every user action
•Checks all 10 principles (not just the obvious ones)
•Provides specific, actionable fixes (not vague suggestions)
•Identifies at least one automation opportunity if any exist
•Ranks findings by impact
•Uses plain language in the review itself (practicing what it preaches) </success_criteria>