Get a check's full review
get_reviewCall this before telling the person a change works, before closing its ticket, and when you are going to fix what the check found — it is the evidence, so the person does not have to look. The run's result in the shape you act on. Every finding in full (where it happens, what was tried, what happened, why it matters, evidence URLs — the frame of the step that saw it first), every journey step this check walked with its screenshot_url (open it to show the person what the check saw), what was not covered, and per finding the sentence that says when it counts as gone (a journey with carried: true was walked by an earlier check and lists no steps: get_review its carried_from) (next_actions). A journey that checks a scenario with words of its own carries expectations — each sentence that says something must be seen, with result seen / wrong / not_seen, why it was not seen, and the number of the step that shows it — and outcome: confirmed (every one seen), not_confirmed, or broken (one shows otherwise). It names symptoms and evidence, never files or fixes — what to change is your call. Name the check by run_id, or by deploy_sha: the commit a check was bound to, also when the GitHub App or the Action started it and you have no run id. By deploy_sha alone it is the commit's latest finished check, whoever started it; started_by: "github" picks the GitHub App's own check of that deploy and says if it is still running (code not_finished, with run_id and status), was refused (deploy_refused, with refusal) or never reached the App (no_deploy_received). run.started_by and run.narrow say who started a check and whether it walked only named journeys. changed_screens: the steps that look different from the previous check of the app (or from compare_to), each with before_url and after_url — after a UI change, show these to the person instead of taking screenshots yourself; what is broken stays in the findings.
Tool call template
This tool declares no required parameters. Add optional parameters from the input schema as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {},
"name": "get_review"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| team | No | Act in another of your teams, by its name or id (whoami lists them). Omitted: the key's own team | |
| app_id | No | With deploy_sha: which app, when the same commit was checked for several | |
| run_id | No | Run id returned by start_check or latest_results, or the run's number as the verdict page shows it ("#332") | |
| compare_to | No | The check to compare screens with (changed_screens), e.g. the app's last production check for a PR preview; default: the previous check of the same app | |
| deploy_sha | No | Or: the commit a check was bound to (deploy_sha of start_check, the Action, the GitHub App) — its latest finished check | |
| started_by | No | With deploy_sha: only the check this door started. github = the GitHub App's check of that deploy — finished, still running, refused, or none received |