Add Persona Walkthrough Specialist

[hedonnn]

What does this PR do?

Adds a new Persona Walkthrough Specialist agent to design/ — a UX research agent that simulates cognitive walkthroughs of web pages from the psychological perspective of defined user personas.

Instead of running generic usability checklists, this agent becomes a specific user: it adopts their fears, cultural expectations, search intent, and decision style, then experiences the page scroll by scroll through concurrent think-aloud monologues paired with structured framework analysis.

The approach is grounded in established UX research methodology (cognitive walkthroughs, think-aloud protocols, five-second testing) and validated by recent academic work — notably [UXAgent](https://arxiv.org/html/2504.09407v2) (Amazon Science, 2025), which demonstrated that LLM agents conditioned on personas can produce meaningful usability insights on real websites.

Agent Information

• Agent Name: Persona Walkthrough Specialist
• Category: design
• Specialty: Persona-driven cognitive walkthroughs with CRO framework analysis (LIFT, Cialdini, Fogg)

What makes this agent different

Most UX evaluation agents run against heuristic checklists. This one produces two distinct voices per page fold:

1. The Persona — a raw, first-person inner monologue that sounds like a real person scanning a page ("I have no idea who these people are. Is this an agency? A marketplace?"), not a consultant reporting ("The value proposition lacks clarity").

2. The Analyst — a structured assessment mapping the persona's reaction to conversion frameworks: which LIFT factor is affected, which Cialdini principle is active or missing, where the persona sits on the Fogg Motivation/Ability/Prompt axes.

The agent includes:

• A detailed persona profile template with psychological depth (Bowlby attachment theory, decision style, cultural trust patterns)
• Four structured walkthrough phases (pre-arrival → five-second test → progressive scroll → verdict)
• Prioritized recommendation templates tied to specific folds and framework principles
• Advanced capabilities: multi-persona comparison, competitive walkthroughs, cross-cultural adaptation, longitudinal tracking
• A full framework quick-reference section (LIFT Model, Cialdini's 7 Principles, Fogg Behavior Model)

Motivation

Landing page evaluation is one of the most common tasks where AI agents can add real value — but generic "review this page" prompts produce generic output. The gap is empathy: understanding what a specific person with specific fears and expectations actually experiences when they land on a page.

This agent fills that gap by combining persona psychology with conversion science, producing insights that analytics dashboards and heuristic checklists miss — the human friction that kills conversions.

Testing

Tested on real estate, SaaS, and local service landing pages with multiple persona profiles (varying urgency levels, attachment styles, and cultural backgrounds). The dual-voice format (persona monologue + analyst assessment) consistently produces more actionable insights than single-voice approaches.

Checklist

• Follows agent template structure from CONTRIBUTING.md
• Includes YAML frontmatter with name, description, color, emoji, vibe
• Has concrete template examples (Persona Profile, Analyst Assessment, Verdict, Recommendation)
• Includes personality and voice (with good/bad monologue examples)
• Defines success metrics
• Includes step-by-step workflow (5 phases)
• Proofread and formatted correctly
• Tested in real scenarios

[bensl84]

Closing — this is my personal fork (bensl84/agency-agents). The canonical project lives at https://github.com/msitarzewski/agency-agents — please re-open your PR there. Thanks!