Build an agent that checks a RenX user journey and verifies fixes
Project brief
Build a runnable agent workflow that walks one RenX user journey in a test environment, spots predefined experience failures and repeats the journey after a fix. For example, check whether a user can open a task, see required information and reach the proposal step. We will agree one journey of up to five steps and a specific test surface before contracting. This is automated experience checking, not UI redesign or an opinion report.
The pilot must detect three injected problems: a broken primary action, missing required information, and a stalled/error state that prevents completion. On detection, record the failed step, expected versus observed behavior, a screenshot or equivalent evidence and reproduction steps. It must complete the healthy journey successfully and avoid reporting a defect on that healthy fixture.
After we switch the fixtures to their corrected state, the workflow must rerun and show that each previously reported issue is resolved. Use test identities and synthetic data only. Stop before any real publish, purchase, send or destructive action; completion boundaries are part of the approved test definition. Evidence must exclude tokens and unrelated personal data.
Deliver executable workflow/configuration and instructions that RenX can run again. Human preference and visual taste are not claimed as automatically proven. The result is a repeatable way to catch concrete journey failures and verify their correction.
Delivery deadline: 2026-09-29. Planned work window: up to 10 calendar days after contracting and receipt of the agreed inputs, within this deadline. Confirm a feasible start date, exact scope and required materials before contract acceptance. If a later start needs a different deadline, agree and record the revised deadline before accepting; no work is requested before agreement. RenX supplies the agreed authorized or synthetic materials before work begins. Additional tools, model usage and other third-party costs require separate agreement.
Deliverables & acceptance
What you'll deliver
- Runnable agent workflow for one agreed journey of up to five steps.
- Structured failure records with sanitized evidence and reproduction steps.
- Healthy-run, three failure-run and corrected-run results.
What the result must meet
- The workflow completes the healthy journey and detects all three predefined injected failures at the correct step.
- Each detected issue includes expected/observed behavior, reproduction steps and inspectable evidence.
- Reruns on corrected fixtures confirm resolution of all three issues without reporting failures on the healthy fixture.
- Runs stay within the agreed test boundary and do not publish, purchase, message real users or expose sensitive information.