Skip to main content

run_tests

Run the project’s saved checks on simulators; returns the summary and JUnit/HTML reports, or a suite_id to wait on. With pr, test what this branch changed and answer with the verdict, frames and GIFs. Acts on the device. Arguments Example
tools/call params
Returns While the suite runs past timeout_s, a line with its progress, and structuredContent has status running, suite_id, done, total and message: call it again with suite_id to keep waiting. When it finishes, the counts, each check that didn’t pass and why, healed steps, and the report and JUnit paths. structuredContent has status finished, suite_id, exit_code, summary, failures, report, junit and results. With pr, it has headline, counts, rows, drafts, comment, pr_json and report in place of summary, failures, junit and results, and up to 2 failed checks’ frames come as JPEGs.
Annotations: title: "Run tests", readOnlyHint: false, destructiveHint: false, openWorldHint: true

test_history

Read the project’s test history: each check’s last 20 results per device, pass rate and flake score, flaky first. Read-only. Arguments Example
tools/call params
Returns One line per check and device: its state and how many runs passed. structuredContent has ok (false when the history file couldn’t be read), total and checks.
Annotations: title: "Test history", readOnlyHint: true, destructiveHint: false, idempotentHint: true, openWorldHint: false

test_schedule

list: the project’s nightly run, when it runs and last night’s verdict. run: run it now on simulators; a schedule that names an iPhone is refused. Local. Arguments Example
tools/call params
Returns With list, whether the nightly run is on, its time and flags, and its last run: structuredContent has on, at, flags and last. With run, what run_tests returns.
Annotations: title: "Test schedule", readOnlyHint: false, destructiveHint: false, openWorldHint: false

review_heals

List the healed steps waiting for review (each with the step before and after, and frames), or settle them: accept keeps them; reject restores the recorded steps. A heal never changes a check’s expectations. Local. Arguments Example
tools/call params
Returns With list, each healed step waiting for review: its check, step and what it became. structuredContent has heals, each with check, diff and heals. With accept or reject, the checks it settled: structuredContent has action and settled, each with check and steps.
Annotations: title: "Review heals", readOnlyHint: false, destructiveHint: false, openWorldHint: false

import_maestro

Turn Maestro flows into checks in the project’s .mobster/checks. A check that exists is kept unless force; dry_run writes nothing; a flow Mobster can’t run is refused. Local. Arguments Example
tools/call params
Returns What it converted and wrote, flow by flow. structuredContent has flows, folder, written, refused and dry_run.
Annotations: title: "Import maestro", readOnlyHint: false, destructiveHint: false, openWorldHint: false

repro_crash

Turn a crash (.ips, Sentry event JSON, TestFlight or Xcode crash log, or a brief) into a check that reproduces it on a simulator, recorded. Run it after your fix: it passes once the app stops crashing. Local. Arguments Fields of brief Fields of brief.frames[] Example
tools/call params
Returns While the suite runs past timeout_s, a line with its progress, and structuredContent has status running, suite_id, done, total and message: call it again with suite_id to keep waiting. When it finishes, whether the check reproduced the crash, with the steps it tried. structuredContent has status, suite_id, reproduction, brief, check, stepsTried, crash, video, report and note.
Annotations: title: "Repro crash", readOnlyHint: false, destructiveHint: false, openWorldHint: false

compare_test_suites

Compare two exact completed local test suites. Returns both verdicts and bounded case transitions; never treats a green run alone as proof of a fix. Read-only. Arguments Example
tools/call params
Returns Whether the suites are comparable, whether the candidate verifies a fix, and how many case and target pairs it shows. structuredContent has schema, baseSuite, candidateSuite, baseVerdict, candidateVerdict, comparable, verifiedFix, proof, reasons, cases (at most 100), casesTotal and casesTruncated.
Annotations: title: "Compare test suites", readOnlyHint: true, destructiveHint: false, idempotentHint: true, openWorldHint: false