Skip to main content
A run report is the per-story view: one user story, one build, one device. Open it from the run view by clicking any row. Run report — left rail with per-story verdicts, right pane with criterion detail and video timeline

​
Verdict header

  • Passed — every critical criterion held.
  • Warning — every critical criterion held, at least one warning criterion didn’t.
  • Failed — at least one critical criterion didn’t hold.
  • Unprocessable — Mini ran the story but couldn’t grade it.
The first three are normal outcomes you triage. Unprocessable is the one that means something is wrong before the test even gets a chance — see below.

​
Criterion list

Each criterion appears with its status and the agent’s evidence:
  • ✅ Passed — what the agent saw that matched.
  • ❌ Failed — what was expected vs what the agent saw, with a screenshot pinned to the moment of failure.
  • ⚠️ Warning — the criterion is non-critical and didn’t hold.
Click any criterion to jump to that moment in the video.

​
Video + timeline

A continuous recording of the agent driving the app. The timeline underneath marks each criterion’s start/end, so you can scrub straight to the failure.

​
Fix prompt

Every failed criterion exposes a Copy fix prompt button. The prompt is a ready-to-paste text block Mini wrote while running the story, containing three things:
  1. The root cause — what went wrong and why.
  2. Steps to reproduce — the path the agent took into the failure.
  3. A concrete proposed fix — a starting point, not a final answer.
Paste it into Cursor or Claude Code. Your IDE has the rest of your codebase in context, so the fix prompt plus that context is usually enough for the agent to open a PR. Run report — Prompt to fix tab on a failed criterion with the Copy for Cursor / Claude button The fix prompt is deliberately prose — no screenshot URLs, no log dumps. The video and the criterion detail above already cover the evidence; the fix prompt is what you hand the IDE.

​
Build fix prompts

When a build itself fails to compile or install — distinct from a story failing on a healthy build — the build status panel exposes its own Copy fix prompt button. Same shape: root cause, repro, proposed fix. Use it when the run shows up as Unprocessable because the build never made it onto a device.

​
When the verdict is Unprocessable

Unprocessable means Mini ran the story but couldn’t produce a real grade — every acceptance criterion ended up either un-evaluable or blocked by something upstream. The header explains which case it was. The common ones:
  • The build is broken — won’t install, crashes on launch, or the bundle is malformed. Open the build panel and use its fix prompt.
  • Sign-in failed — Mini couldn’t get past auth, so nothing downstream is testable. Attach a working profile to the user story.
  • The criteria don’t fit the app — usually a story copied from another app, or a criterion that asks for something the app doesn’t do. Edit the story.
  • A hard blocker mid-story — something earlier in the flow made every remaining criterion unreachable. The first failed criterion is the one to read.
If you see Unprocessable repeatedly across stories on the same build, suspect the build before the suite.