Test a user journey
SkillWeb & browsingWith this skill, your AI can attempt a task on a website the way a real visitor would and tell you whether the site let it succeed. You give it a start URL and a short brief describing the goal, and it works through the pages in a browser without any inside knowledge of how the site is built. It comes back with the route it took, where it hit friction, and the outcome it could actually observe.
Available today. Use it from your connected AI after setup.
No other account needed.
After adding it, give your AI a start URL and a brief that describes the task a user should be able to complete, and run it in an isolated tester context. It will walk the site and return the path, friction, and result.
Then ask your AI: use the Test a user journey skill
What your AI can do with it
- Work through a website in a browser to complete a goal you describe in a brief
- Report the exact path it took through the interface
- Point out where the experience was confusing, slow, or blocked
- State the observable outcome, including when the goal could not be completed
- Run against a start URL you supply, in an isolated tester context kept separate from your live site
What this skill tells your AI
The instructions your AI receives, as published by emdash-cms/emdash in acceptance/tester/SKILL.md and read by ahel’s review.
Act as the person described in the tester brief. Make decisions from the rendered interface and the information that persona would know.
Boundaries
- Use the available browser controls and visible interface.
- Do not inspect a source repository or read files outside the tester workspace.
- Do not call application APIs, query the database, execute JavaScript in the page, inspect the DOM directly, or use developer tools to discover the intended route.
- Do not rely on remembered EmDash routes, control labels, or implementation details.
- Do not change code or suggest a fix during the journey.
The start URL may pass through a local authentication page. Begin judging the journey after the browser reaches the intended starting surface.
Run the journey
Observe the full page and its accessible controls before acting. Use screenshots as well as accessibility information when both are available.
Take one meaningful, user-visible action at a time. After each action, record:
- what you tried;
- what you expected;
- what happened;
- any uncertainty or surprise.
Choose the next action from the visible result. Do not follow an imagined ideal path. Try a reasonable recovery when the interface suggests one, but do not repeat an action that has already failed twice.
Capture evidence at the starting state, important transitions, any failure or confusing state, and the final state. Record user-visible delays or instability, but do not treat local development compilation time as product performance unless the brief asks you to assess it.
Stop when the goal is visibly complete, the interface prevents further progress, or the brief's limit is reached.
Report
Return a Markdown report with these sections:
# UX acceptance report
## Outcome
Completed, gave up, blocked, or uncertain. Describe the visible final state.
## Path taken
Number each meaningful action and its result.
## Friction and observations
For each finding, identify the action, what was confusing or difficult, its user impact, and the supporting screenshot.
## Positive signals
Record feedback or controls that materially helped complete the goal. Omit this section when there were none.
## Evidence
List screenshots and any user-visible error text.
Do not convert uncertainty into a pass. The coordinator will verify the saved state independently.
Signals
- GitHub stars
- 12k
- Forks
- 1k
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
ux-acceptance-tester- Source
- github.com/emdash-cms/emdash