name: "usability-testing"
description: "Plan, run, and synthesize usability tests and produce a Usability Test Pack (test plan, tasks/script, logistics, notes template, issue log, findings + recommendations). Use for usability test, user test, prototype test, user testing, usability study, Wizard of Oz, fake door."

Usability Testing

Scope

Covers
- Designing task-based usability studies tied to a specific product decision
- Testing live flows, prototypes, and “faked” implementations (fake door, Wizard of Oz)
- Running moderated sessions (remote or in-person) and capturing high-quality evidence
- Turning findings into a prioritized fix list (including high-ROI microcopy/CTA improvements)

When to use
- “Create a usability test plan and script for .”
- “We need to test a prototype with 5–8 users next week.”
- “Validate a value proposition before building (fake door / Wizard of Oz).”
- “Help me synthesize usability findings into a prioritized backlog.”

When NOT to use
- You need statistically reliable estimates or causal impact (use analytics/experimentation)
- You need open-ended discovery (“what problems do users have?”) → use conducting-user-interviews
- You’re working with high-risk populations or sensitive topics (medical, legal, minors) without appropriate approvals/training
- You don’t have a concrete scenario/flow to evaluate (clarify the decision first)

Inputs

Minimum required
- Product + target user segment (who, context of use)
- The decision this test should inform (what will change) + timeline
- What you’re testing (flow/feature) + prototype/build link (or “recommend stimulus”)
- Platform + environment (web/mobile/desktop; remote/in-person)
- Constraints: session type, number of participants, incentives, recording policy, privacy constraints

Missing-info strategy
- Ask up to 5 questions from references/INTAKE.md.
- If still unknown, proceed with explicit assumptions and list Open questions that would change the plan.

Outputs (deliverables)

Produce a Usability Test Pack in Markdown (in-chat; or as files if requested):

1) Context snapshot (decision, users, what’s being tested, constraints)
2) Test plan (method, prototype strategy, hypotheses/risks, success criteria)
3) Participant plan (criteria, recruiting channels, schedule + backups)
4) Moderator guide + task script (neutral tasks, probes, wrap-up)
5) Note-taking template + issue log (severity/impact, evidence)
6) Synthesis readout (findings, prioritized issues, recommendations, quick wins)
7) Risks / Open questions / Next steps (always included)

Templates: references/TEMPLATES.md
Expanded heuristics: references/WORKFLOW.md

Workflow (8 steps)

1) Frame the decision and the “why now”

• Inputs: User context; references/INTAKE.md.
• Actions: Define the decision, primary unknowns, and the minimum you need to learn to make the call.
• Outputs: Context snapshot + research questions/hypotheses.
• Checks: You can answer: “What will we do differently after this test?”

2) Choose the right stimulus (real vs prototype vs faked)

• Inputs: What’s being tested; constraints.
• Actions: Select the cheapest valid setup: live product, clickable prototype, fake door, Wizard of Oz, or concierge flow.
• Outputs: Prototype strategy + what will be real vs simulated.
• Checks: The setup tests the core value/behavior (not pixel perfection).

3) Define tasks and success criteria (keep it neutral)

• Inputs: User goals + scenarios.
• Actions: Write 5–8 realistic tasks (each with a starting state), success criteria, and key observables (hesitation, errors, workarounds).
• Outputs: Task list (draft) + observation plan.
• Checks: Tasks don’t reveal UI labels (“Click the X button”); they reflect real intent.

4) Pick participants + recruiting plan (include buffers)

• Inputs: Target segment, access to users.
• Actions: Set inclusion/exclusion criteria; choose channels; build a schedule with backups and slack for no-shows and busy participants.
• Outputs: Participant plan + recruiting copy/screener (as needed).
• Checks: Participants match the scenario (behavior/context), not just demographics.

5) Build the moderator guide + instrumentation

• Inputs: Task list + prototype.
• Actions: Create the script (intro/consent, warm-up, tasks, probes, wrap-up). Assign note-taker roles; decide what to record.
• Outputs: Moderator guide + notes template + issue log.
• Checks: The guide avoids leading questions and includes “what would you do next?” probes.

6) Run sessions and capture evidence (optional “reality checks”)

• Inputs: Guide, logistics, participants.
• Actions: Run sessions; capture verbatims, errors, rough time-on-task, and moments of confusion. Optionally observe comparable flows “in the wild.”
• Outputs: Completed notes per session + populated issue log.
• Checks: Every issue has at least one concrete example (quote/screenshot/time/step) attached.

7) Synthesize into prioritized fixes (micro wins count)

• Inputs: Notes + issue log.
• Actions: Cluster issues; label severity and frequency; connect to funnel/business impact; propose fixes (including microcopy/CTA tweaks).
• Outputs: Synthesis readout + prioritized recommendations/backlog.
• Checks: Each recommendation ties to evidence and an expected impact (directional).

8) Share, decide, and run the quality gate

• Inputs: Draft pack.
• Actions: Produce a shareable readout, propose next steps (design iteration, follow-up test, experiment). Run references/CHECKLISTS.md and score references/RUBRIC.md.
• Outputs: Final Usability Test Pack + Risks/Open questions/Next steps.
• Checks: A stakeholder can make a “ship / fix / retest” decision asynchronously.

Quality gate (required)

• Use references/CHECKLISTS.md and references/RUBRIC.md.
• Always include: Risks, Open questions, Next steps.

Examples

Example 1 (Prototype test): “Create a usability test plan + moderator guide to evaluate our new onboarding flow (web) with 6 first-time users next week.”
Expected: full Usability Test Pack with neutral tasks, recruiting criteria, session logistics, and a synthesis structure.

Example 2 (Wizard of Oz): “We want to test an ‘AI auto-triage’ feature before building it. Design a Wizard of Oz usability test plan and script for 5 sessions.”
Expected: stimulus plan defining what’s simulated, tasks focused on value, and an issue log + readout.

Boundary example: “Run a usability test to prove the redesign will increase retention by 10%.”
Response: explain limits of small-n usability; recommend pairing with instrumentation/experimentation for causality and use usability to diagnose friction.