fix: address Copilot review feedback on #386

- Add Type field to synthesized _checkpoint_error GraderResults
- Fix docs to reference 'graders:' (the actual YAML key) instead of 'validators:'
- Update schema-changes.md Policy section to reflect current 1.1 default emission while preserving 1.0 reader fallback for back-compat

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: Copilot

.impeccable/hook.cache.json

@@ -0,0 +1 @@
+{"version":1,"sessions":{}}

internal/orchestration/checkpoints.go

@@ -34,8 +34,9 @@ type checkpointRunner struct {
 }
 
 // newCheckpointRunner builds a checkpointRunner from a TestCase. Returns nil
-// when the task has no checkpoints — callers should always check for nil
-// before invoking methods.
+// when the task has no checkpoints, but all methods on *checkpointRunner are
+// nil-safe — callers can invoke runForTurn/results on the returned value
+// directly without a nil guard.
 func newCheckpointRunner(runner *EvalRunner, tc *models.TestCase) *checkpointRunner {
 	if tc == nil || len(tc.Checkpoints) == 0 {
 		return nil
@@ -112,6 +113,7 @@ func (cr *checkpointRunner) runForTurn(ctx context.Context, turn int, resp *exec
 		}
 		outcome.Validations["_checkpoint_error"] = models.GraderResults{
 			Name:     "_checkpoint_error",
+			Type:     models.GraderKind("checkpoint_error"),
 			Score:    0,
 			Passed:   false,
 			Feedback: err.Error(),

internal/orchestration/runner.go

@@ -1231,13 +1231,20 @@ func (r *EvalRunner) executeRun(ctx context.Context, tc *models.TestCase, runNum
 
 	// Surface checkpoint failures in the run status even when graders are
 	// skipped or when the final-pass graders all passed. A failed checkpoint
-	// without on_failure: stop should still mark the run as failed.
+	// without on_failure: stop should still mark the run as failed; a
+	// checkpoint that recorded StatusError (grader-execution error) should
+	// promote the run to StatusError so consumers can distinguish
+	// infrastructure problems from assertion failures.
 	checkpointOutcomes := cps.results()
 	if status != models.StatusError {
 		for _, co := range checkpointOutcomes {
-			if co.Status != models.StatusPassed {
-				status = models.StatusFailed
-				break
+			switch co.Status {
+			case models.StatusError:
+				status = models.StatusError
+			case models.StatusFailed:
+				if status != models.StatusError {
+					status = models.StatusFailed
+				}
 			}
 		}
 	}

site/src/content/docs/guides/eval-yaml.mdx

@@ -443,7 +443,7 @@ Each checkpoint accepts:
 | Field        | Type     | Description                                                                       |
 | ------------ | -------- | --------------------------------------------------------------------------------- |
 | `after_turn` | int      | 1-based turn number this checkpoint runs after (initial prompt is turn 1).        |
-| `graders`    | array    | Inline graders, same schema as task `validators:` / eval-level `graders:`.        |
+| `graders`    | array    | Inline graders, same schema as the task-level `graders:` / eval-level `graders:` field. |
 | `on_failure` | string   | `continue` (default) or `stop` — abort remaining turns when this checkpoint fails.|
 
 Outcomes are recorded per-checkpoint on `results.json` under `checkpoints[]`, alongside the final `validations`. `waza gate` still uses final-pass status. Available with `schemaVersion: "1.1"` and above (additive — older 1.0 files load unchanged).

site/src/content/docs/reference/schema-changes.md

@@ -20,8 +20,8 @@ Schema versions use `MAJOR.MINOR` format with no patch component.
 
 - **MINOR** changes are backward-compatible additions, usually optional fields. Readers accept same-major artifacts and warn when they see unknown fields.
 - **MAJOR** changes are breaking. Readers refuse artifacts from a different major version and point to `waza migrate <file>`.
-- Missing `schemaVersion` defaults to `1.0` for backward compatibility with existing `eval.yaml` and `results.json` files.
-- New artifacts should include `schemaVersion: "1.0"` or `"schemaVersion": "1.0"` explicitly.
+- Missing `schemaVersion` defaults to `1.0` for backward compatibility with eval suites authored before the field was introduced. Same-major readers continue to accept `1.0`-defaulted files unchanged.
+- New artifacts should emit the current `schemaVersion` (currently `1.1`). The version is automatically populated by the writer; you only need to set it manually when authoring fixtures or schema-pinned test data.
 
 ## Migration command
 

site/src/content/docs/reference/schema.mdx

@@ -574,7 +574,7 @@ Per-turn graders that run at specific turn boundaries during multi-turn conversa
 | Field        | Type    | Required | Description                                                                                  |
 |--------------|---------|----------|----------------------------------------------------------------------------------------------|
 | `after_turn` | integer | yes      | 1-based turn number to grade after. Turn 1 is the initial prompt; follow-ups are 2, 3, ...   |
-| `graders`    | array   | yes      | Inline graders (same schema as task `validators:`).                                          |
+| `graders`    | array   | yes      | Inline graders (same schema as the task-level `graders:` field).                             |
 | `on_failure` | string  | no       | `continue` (default) or `stop`. `stop` aborts remaining turns when this checkpoint fails.    |
 
 ```yaml