* Display label for a validation / dry-run target: "main" for the main env, and * "<name> #<pr>" for a PR (with a "(SDK PR)" marker on the auto-detec

* The recipe's `create` graph (the seed request: which entities to create, with their fields/refs), as a * JSON string ready for the diagnoser pr

* Get the current URL.

The recent run history for one test (by app + test slug), newest first.

* Lists the pods for a label selector, returning [] on error. The readiness * loops treat "could not read pods" as "keep waiting", never as a ter

* SDK capability: list the preview envs the dry-run can target (open-PR * previews + main), flagging the auto-detected SDK implementation PR.

* Healing-scope sibling of {@link loadSnapshot}: gather the unified diff-job * context for one refinement iteration's failing subjects. The heali

* Loads the pending snapshot for a branch, verifying it exists and is still pending. * * @throws {ApplicationNotFoundError} If the branch do

* Snapshot-scope sibling of {@link load}: gather the diff-job context across * every replayed, flagged run in a snapshot. Resolution reasons over

* Load the state for an operation that should work from `minimumStep` or any later step. * * If the current step is at or past `minimumStep`

* Best-effort lifecycle write for the PR entry point (running / stage / failed). Fire-and-forget: a progress * write must never sink the run, so both

* Mark the start of a new build attempt for this environment. Successive * attempts (reruns, new commits) share one retention-bounded stream, so

* Navigate to the specified URL. * @param url The URL to navigate to

Splits a multiline field into trimmed, non-empty lines (one argv token / port per line).

Press a given key on the keyboard

Appends a commit to a branch. For the default branch, pass its name (e.g. "main").

* Create a scenario with an active recipe version so that * `WorkingState.startScenarioDryRun()` passes its recipe-count check.

* Creates a branch with an active snapshot that has a test case assigned. * Returns the branchId and testCaseId for use in tests.

The human-readable reason to surface for a scenario-setup failure, if any.

* Deploy path for pull_request opened/synchronize/reopened: starts the deploy * workflow directly. Silently skipped when previews are disabled (dev /

* Maps a terminal failure to its outcome bucket. A scenario-setup failure means * the test never got a chance to run (the environment never came up),

A model that returns free-text on the tool-loop call, then the structured decision on the Output.object call.