diff --git a/apps/cli/src/commands/validate/validate-files.ts b/apps/cli/src/commands/validate/validate-files.ts
index b00c27b47..f9a5021ba 100644
--- a/apps/cli/src/commands/validate/validate-files.ts
+++ b/apps/cli/src/commands/validate/validate-files.ts
@@ -5,6 +5,7 @@ import {
   type ValidationResult,
   type ValidationSummary,
   detectFileType,
+  validateCasesFile,
   validateConfigFile,
   validateEvalFile,
   validateFileReferences,
@@ -17,12 +18,7 @@ import fg from 'fast-glob';
  */
 export async function validateFiles(paths: readonly string[]): Promise<ValidationSummary> {
   const filePaths = await expandPaths(paths);
-  const results: ValidationResult[] = [];
-
-  for (const filePath of filePaths) {
-    const result = await validateSingleFile(filePath);
-    results.push(result);
-  }
+  const results = await Promise.all(filePaths.map((filePath) => validateSingleFile(filePath)));
 
   const validFiles = results.filter((r) => r.valid).length;
   const invalidFiles = results.filter((r) => !r.valid).length;
@@ -58,10 +54,27 @@ async function validateSingleFile(filePath: string): Promise<ValidationResult> {
         };
       }
     }
+  } else if (fileType === 'cases') {
+    result = await validateCasesFile(absolutePath);
   } else if (fileType === 'targets') {
     result = await validateTargetsFile(absolutePath);
-  } else {
+  } else if (fileType === 'config') {
     result = await validateConfigFile(absolutePath);
+  } else {
+    // Unknown file type — skip validation, report as skipped
+    result = {
+      valid: true,
+      filePath: absolutePath,
+      fileType: 'unknown',
+      errors: [
+        {
+          severity: 'warning',
+          filePath: absolutePath,
+          message:
+            'File type not recognized. Eval files must end in .eval.yaml. Skipping validation.',
+        },
+      ],
+    };
   }
 
   return result;
@@ -130,7 +143,7 @@ async function findYamlFiles(dirPath: string): Promise<readonly string[]> {
         }
         const subFiles = await findYamlFiles(fullPath);
         results.push(...subFiles);
-      } else if (entry.isFile() && isYamlFile(entry.name)) {
+      } else if (entry.isFile() && isEvalYamlFile(entry.name)) {
         results.push(fullPath);
       }
     }
@@ -145,3 +158,9 @@ function isYamlFile(filePath: string): boolean {
   const ext = path.extname(filePath).toLowerCase();
   return ext === '.yaml' || ext === '.yml';
 }
+
+/** Returns true only for *.eval.yaml / *.eval.yml files (used for directory scanning). */
+function isEvalYamlFile(filePath: string): boolean {
+  const lower = path.basename(filePath).toLowerCase();
+  return lower.endsWith('.eval.yaml') || lower.endsWith('.eval.yml');
+}
diff --git a/examples/features/basic-jsonl/evals/dataset.eval.yaml b/examples/features/basic-jsonl/evals/dataset.eval.yaml
index c226536db..508c684c0 100644
--- a/examples/features/basic-jsonl/evals/dataset.eval.yaml
+++ b/examples/features/basic-jsonl/evals/dataset.eval.yaml
@@ -7,6 +7,4 @@ name: basic-jsonl
 execution:
   target: llm
 
-evaluator: llm_grader
-
 tests: ./dataset.jsonl
diff --git a/examples/features/prompt-template-sdk/evals/dataset.eval.yaml b/examples/features/prompt-template-sdk/evals/dataset.eval.yaml
index 6917de1bd..81e221837 100644
--- a/examples/features/prompt-template-sdk/evals/dataset.eval.yaml
+++ b/examples/features/prompt-template-sdk/evals/dataset.eval.yaml
@@ -18,8 +18,9 @@ tests:
           - type: text
             value: What are the main benefits of TypeScript over JavaScript?
 
-    reference_answer: |-
-      TypeScript provides static type checking, better IDE support, and improved maintainability.
+    expected_output:
+      - role: assistant
+        content: TypeScript provides static type checking, better IDE support, and improved maintainability.
 
     assertions:
       - name: custom-prompt-eval
@@ -37,8 +38,9 @@ tests:
           - type: text
             value: Explain async/await in JavaScript.
 
-    reference_answer: |-
-      Async/await is syntactic sugar over Promises that makes asynchronous code look synchronous.
+    expected_output:
+      - role: assistant
+        content: Async/await is syntactic sugar over Promises that makes asynchronous code look synchronous.
 
     assertions:
       - name: strict-eval
diff --git a/examples/features/tool-trajectory-advanced/evals/trace-file-demo.eval.yaml b/examples/features/tool-trajectory-advanced/evals/trace-file-demo.eval.yaml
index 07609020a..a978cee75 100644
--- a/examples/features/tool-trajectory-advanced/evals/trace-file-demo.eval.yaml
+++ b/examples/features/tool-trajectory-advanced/evals/trace-file-demo.eval.yaml
@@ -23,10 +23,8 @@ tests:
   # Expected score: 1.0 (webSearch before fetchPage is satisfied)
   # =============================================================================
   - id: in-order-validation
-    description: |-
-      Validates that the agent performs web search before fetching page details.
-      Mode 'in_order' allows other tool calls between expected tools.
-
+    # Validates that the agent performs web search before fetching page details.
+    # Mode 'in_order' allows other tool calls between expected tools.
     criteria: |-
       Agent searches for product information, then fetches detailed specs.
 
@@ -49,10 +47,8 @@ tests:
   # Expected score: 1.0 (matches full trace exactly)
   # =============================================================================
   - id: exact-sequence-validation
-    description: |-
-      Validates the exact sequence of all tool calls in the trace.
-      Mode 'exact' requires the trace to match precisely.
-
+    # Validates the exact sequence of all tool calls in the trace.
+    # Mode 'exact' requires the trace to match precisely.
     criteria: |-
       Agent follows the exact research workflow: search, fetch, search reviews, summarize.
 
@@ -76,10 +72,8 @@ tests:
   # Expected score: 1.0 (meets all minimums)
   # =============================================================================
   - id: any-order-with-minimums
-    description: |-
-      Validates that the agent performs adequate research by checking minimum
-      tool call counts. Mode 'any_order' with minimums is flexible on sequence.
-
+    # Validates that the agent performs adequate research by checking minimum
+    # tool call counts. Mode 'any_order' with minimums is flexible on sequence.
     criteria: |-
       Agent performs at least 2 web searches and 1 page fetch for thorough research.
 
@@ -101,10 +95,8 @@ tests:
   # Expected score: 1.0 (inputs match expected patterns)
   # =============================================================================
   - id: tool-input-validation
-    description: |-
-      Validates that tool calls include appropriate input parameters.
-      Useful for ensuring the agent provides correct context to tools.
-
+    # Validates that tool calls include appropriate input parameters.
+    # Useful for ensuring the agent provides correct context to tools.
     criteria: |-
       Agent searches with relevant product keywords and fetches from authoritative source.
 
@@ -130,10 +122,8 @@ tests:
   # Expected score: 1.0 (outputs contain expected fields)
   # =============================================================================
   - id: tool-output-validation
-    description: |-
-      Validates that tool outputs contain expected data.
-      Useful for regression testing when tool behavior changes.
-
+    # Validates that tool outputs contain expected data.
+    # Useful for regression testing when tool behavior changes.
     criteria: |-
       Web search returns results with links and snippets; fetch returns product specs.
 
@@ -161,10 +151,8 @@ tests:
   # This mirrors patterns used in complex agent pipelines
   # =============================================================================
   - id: combined-validation
-    description: |-
-      Production-style evaluation combining sequence validation with input/output
-      checks. Demonstrates a realistic multi-turn agent workflow.
-
+    # Production-style evaluation combining sequence validation with input/output
+    # checks. Demonstrates a realistic multi-turn agent workflow.
     criteria: |-
       Agent performs comprehensive product research:
       1. Initial web search for product specs
diff --git a/examples/showcase/tool-evaluation-plugins/README.md b/examples/showcase/tool-evaluation-plugins/README.md
index fe779397a..cbf91eaca 100644
--- a/examples/showcase/tool-evaluation-plugins/README.md
+++ b/examples/showcase/tool-evaluation-plugins/README.md
@@ -56,7 +56,7 @@ evaluators:
 export TOOL_EVAL_PLUGINS_DIR=$(pwd)/examples/showcase/tool-evaluation-plugins
 
 # Run the demo
-npx agentv eval examples/showcase/tool-evaluation-plugins/tool-eval-demo.yaml
+npx agentv eval examples/showcase/tool-evaluation-plugins/tool-eval-demo.eval.yaml
 ```
 
 ## Input Contract
diff --git a/examples/showcase/tool-evaluation-plugins/tool-eval-demo.yaml b/examples/showcase/tool-evaluation-plugins/tool-eval-demo.eval.yaml
similarity index 98%
rename from examples/showcase/tool-evaluation-plugins/tool-eval-demo.yaml
rename to examples/showcase/tool-evaluation-plugins/tool-eval-demo.eval.yaml
index 0dab62823..5ba64e4d2 100644
--- a/examples/showcase/tool-evaluation-plugins/tool-eval-demo.yaml
+++ b/examples/showcase/tool-evaluation-plugins/tool-eval-demo.eval.yaml
@@ -5,7 +5,7 @@
 # semantic evaluation capabilities that require domain-specific logic.
 #
 # Run: cd examples/showcase/tool-evaluation-plugins
-#      npx agentv eval tool-eval-demo.yaml --target mock_agent
+#      npx agentv eval tool-eval-demo.eval.yaml --target mock_agent
 
 description: Showcase of tool evaluation plugin patterns
 
diff --git a/packages/core/src/evaluation/validation/cases-validator.ts b/packages/core/src/evaluation/validation/cases-validator.ts
new file mode 100644
index 000000000..d88a7a232
--- /dev/null
+++ b/packages/core/src/evaluation/validation/cases-validator.ts
@@ -0,0 +1,98 @@
+import { readFile } from 'node:fs/promises';
+import path from 'node:path';
+import { parse } from 'yaml';
+
+import type { ValidationError, ValidationResult } from './types.js';
+
+type JsonValue = string | number | boolean | null | JsonObject | JsonArray;
+type JsonObject = { readonly [key: string]: JsonValue };
+type JsonArray = readonly JsonValue[];
+
+function isObject(value: unknown): value is JsonObject {
+  return typeof value === 'object' && value !== null && !Array.isArray(value);
+}
+
+/**
+ * Validate a cases file — a YAML file whose root is an array of test case objects.
+ *
+ * Cases files are referenced from eval files via `tests: path/to/cases.yaml` or
+ * `file://cases/accuracy.yaml` entries in the tests array. Each item must have
+ * at least an `id` (non-empty string) and an `input` (string or array).
+ */
+export async function validateCasesFile(filePath: string): Promise<ValidationResult> {
+  const errors: ValidationError[] = [];
+  const absolutePath = path.resolve(filePath);
+
+  let parsed: unknown;
+  try {
+    const content = await readFile(absolutePath, 'utf8');
+    parsed = parse(content);
+  } catch (error) {
+    errors.push({
+      severity: 'error',
+      filePath: absolutePath,
+      message: `Failed to parse YAML: ${(error as Error).message}`,
+    });
+    return { valid: false, filePath: absolutePath, fileType: 'cases', errors };
+  }
+
+  if (!Array.isArray(parsed)) {
+    errors.push({
+      severity: 'error',
+      filePath: absolutePath,
+      message: 'Cases file must contain a YAML array of test case objects',
+    });
+    return { valid: false, filePath: absolutePath, fileType: 'cases', errors };
+  }
+
+  for (let i = 0; i < parsed.length; i++) {
+    const item = parsed[i];
+    const location = `[${i}]`;
+
+    if (!isObject(item)) {
+      errors.push({
+        severity: 'error',
+        filePath: absolutePath,
+        location,
+        message: 'Each test case must be an object',
+      });
+      continue;
+    }
+
+    // Required: id
+    const id = item.id;
+    if (typeof id !== 'string' || id.trim().length === 0) {
+      errors.push({
+        severity: 'error',
+        filePath: absolutePath,
+        location: `${location}.id`,
+        message: "Missing or invalid 'id' field (must be a non-empty string)",
+      });
+    }
+
+    // Required: input
+    const input = item.input;
+    if (input === undefined) {
+      errors.push({
+        severity: 'error',
+        filePath: absolutePath,
+        location: `${location}.input`,
+        message: "Missing 'input' field (must be a string or array of messages)",
+      });
+    } else if (typeof input !== 'string' && !Array.isArray(input)) {
+      errors.push({
+        severity: 'error',
+        filePath: absolutePath,
+        location: `${location}.input`,
+        message: "Invalid 'input' field (must be a string or array of messages)",
+      });
+    }
+  }
+
+  return {
+    valid: errors.filter((e) => e.severity === 'error').length === 0,
+    filePath: absolutePath,
+    fileType: 'cases',
+    errors,
+  };
+}
diff --git a/packages/core/src/evaluation/validation/eval-file.schema.ts b/packages/core/src/evaluation/validation/eval-file.schema.ts
index e35f46287..7c6bbc392 100644
--- a/packages/core/src/evaluation/validation/eval-file.schema.ts
+++ b/packages/core/src/evaluation/validation/eval-file.schema.ts
@@ -382,7 +382,6 @@ const EvalTestSchema = z.object({
   metadata: z.record(z.unknown()).optional(),
   conversation_id: z.string().optional(),
   suite: z.string().optional(),
-  note: z.string().optional(),
   depends_on: z.array(z.string()).optional(),
   on_dependency_failure: z.enum(['skip', 'fail', 'run']).optional(),
   mode: z.enum(['conversation']).optional(),
diff --git a/packages/core/src/evaluation/validation/eval-validator.ts b/packages/core/src/evaluation/validation/eval-validator.ts
index 576dc9f78..f22f82e15 100644
--- a/packages/core/src/evaluation/validation/eval-validator.ts
+++ b/packages/core/src/evaluation/validation/eval-validator.ts
@@ -1,4 +1,4 @@
-import { readFile } from 'node:fs/promises';
+import { readFile, readdir } from 'node:fs/promises';
 import path from 'node:path';
 import { parse } from 'yaml';
 
@@ -36,6 +36,7 @@ const KNOWN_TOP_LEVEL_FIELDS = new Set([
   '$schema',
   'name',
   'description',
+  'category',
   'version',
   'author',
   'tags',
@@ -44,14 +45,25 @@ const KNOWN_TOP_LEVEL_FIELDS = new Set([
   'input',
   'input_files',
   'tests',
-  'eval_cases',
   'target',
   'execution',
   'assertions',
   'evaluators',
+  'preprocessors',
   'workspace',
 ]);
 
+/**
+ * Deprecated top-level fields with migration hints.
+ * These are still processed by yaml-parser but authors should migrate.
+ */
+const DEPRECATED_TOP_LEVEL_FIELDS = new Map<string, string>([
+  ['eval_cases', "'eval_cases' is deprecated. Use 'tests' instead."],
+  ['evalcases', "'evalcases' is deprecated. Use 'tests' instead."],
+  ['evaluator', "'evaluator' is deprecated. Use 'assertions' instead."],
+  ['assert', "'assert' is deprecated. Use 'assertions' instead."],
+]);
+
 /** Known fields at the test level. */
 const KNOWN_TEST_FIELDS = new Set([
   'id',
@@ -61,12 +73,12 @@ const KNOWN_TEST_FIELDS = new Set([
   'expected_output',
   'assertions',
   'evaluators',
+  'rubrics',
   'execution',
   'workspace',
   'metadata',
   'conversation_id',
   'suite',
-  'note',
   'depends_on',
   'on_dependency_failure',
   'mode',
@@ -76,9 +88,66 @@ const KNOWN_TEST_FIELDS = new Set([
   'window_size',
 ]);
 
+/**
+ * Deprecated test-level fields with migration hints.
+ * These are still processed by yaml-parser but authors should migrate.
+ */
+const DEPRECATED_TEST_FIELDS = new Map<string, string>([
+  ['evaluator', "'evaluator' is deprecated. Use 'assertions' instead."],
+  ['assert', "'assert' is deprecated. Use 'assertions' instead."],
+  ['expected_outcome', "'expected_outcome' is deprecated. Use 'criteria' instead."],
+]);
+
 /** Name field pattern: lowercase alphanumeric with hyphens. */
 const NAME_PATTERN = /^[a-z0-9-]+$/;
 
+/** Script file extensions recognised as custom assertion plugins. */
+const ASSERTION_SCRIPT_EXTENSIONS = new Set(['.ts', '.js', '.mts', '.mjs', '.cts', '.cjs']);
+
+/** Cache: directory path → promise of discovered type names. */
+const customAssertionCache = new Map<string, Promise<Set<string>>>();
+
+/**
+ * Walk up the directory tree from `baseDir` collecting type names from
+ * `.agentv/assertions/` directories — mirrors the runtime discovery in
+ * `assertion-discovery.ts`.
+ *
+ * Results are cached by directory so concurrent validation of many files
+ * in the same directory only does the filesystem walk once.
+ */
+function discoverCustomAssertionTypes(baseDir: string): Promise<Set<string>> {
+  const resolved = path.resolve(baseDir);
+  const cached = customAssertionCache.get(resolved);
+  if (cached) return cached;
+
+  const promise = (async () => {
+    const types = new Set<string>();
+    let dir = resolved;
+    const root = path.parse(dir).root;
+
+    while (dir !== root) {
+      const assertionsDir = path.join(dir, '.agentv', 'assertions');
+      try {
+        const entries = await readdir(assertionsDir, { withFileTypes: true });
+        for (const entry of entries) {
+          if (!entry.isFile()) continue;
+          const ext = path.extname(entry.name).toLowerCase();
+          if (!ASSERTION_SCRIPT_EXTENSIONS.has(ext)) continue;
+          types.add(entry.name.slice(0, -ext.length));
+        }
+      } catch {
+        // Directory doesn't exist — skip
+      }
+      dir = path.dirname(dir);
+    }
+
+    return types;
+  })();
+
+  customAssertionCache.set(resolved, promise);
+  return promise;
+}
+
 function isObject(value: unknown): value is JsonObject {
   return typeof value === 'object' && value !== null && !Array.isArray(value);
 }
@@ -89,6 +158,7 @@ function isObject(value: unknown): value is JsonObject {
 export async function validateEvalFile(filePath: string): Promise<ValidationResult> {
   const errors: ValidationError[] = [];
   const absolutePath = path.resolve(filePath);
+  const customAssertionTypes = await discoverCustomAssertionTypes(path.dirname(absolutePath));
 
   let parsed: unknown;
   try {
@@ -125,9 +195,17 @@ export async function validateEvalFile(filePath: string): Promise<ValidationResu
   // Validate metadata fields
   validateMetadata(parsed, absolutePath, errors);
 
-  // Warn on unknown top-level fields
+  // Warn on deprecated or unknown top-level fields
   for (const key of Object.keys(parsed)) {
-    if (!KNOWN_TOP_LEVEL_FIELDS.has(key)) {
+    const deprecationMessage = DEPRECATED_TOP_LEVEL_FIELDS.get(key);
+    if (deprecationMessage) {
+      errors.push({
+        severity: 'warning',
+        filePath: absolutePath,
+        location: key,
+        message: deprecationMessage,
+      });
+    } else if (!KNOWN_TOP_LEVEL_FIELDS.has(key)) {
       errors.push({
         severity: 'warning',
         filePath: absolutePath,
@@ -239,9 +317,17 @@ export async function validateEvalFile(filePath: string): Promise<ValidationResu
       continue;
     }
 
-    // Warn on unknown test-level fields
+    // Warn on deprecated or unknown test-level fields
     for (const key of Object.keys(evalCase)) {
-      if (!KNOWN_TEST_FIELDS.has(key)) {
+      const deprecationMessage = DEPRECATED_TEST_FIELDS.get(key);
+      if (deprecationMessage) {
+        errors.push({
+          severity: 'warning',
+          filePath: absolutePath,
+          location: `${location}.${key}`,
+          message: deprecationMessage,
+        });
+      } else if (!KNOWN_TEST_FIELDS.has(key)) {
         errors.push({
           severity: 'warning',
           filePath: absolutePath,
@@ -332,7 +418,7 @@ export async function validateEvalFile(filePath: string): Promise<ValidationResu
     // assertions field (array of assertion objects)
     const assertField = evalCase.assertions;
     if (assertField !== undefined) {
-      validateAssertArray(assertField, location, absolutePath, errors);
+      validateAssertArray(assertField, location, absolutePath, errors, customAssertionTypes);
     }
 
     // Cross-field validation for conversation mode
@@ -612,6 +698,7 @@ function validateAssertArray(
   parentLocation: string,
   filePath: string,
   errors: ValidationError[],
+  customAssertionTypes: ReadonlySet<string> = new Set(),
 ): void {
   if (!Array.isArray(assertField)) {
     errors.push({
@@ -669,7 +756,7 @@ function validateAssertArray(
     // Normalize snake_case to kebab-case for backward compatibility
     const typeValue = rawTypeValue.replace(/_/g, '-');
 
-    if (!isEvaluatorKind(typeValue)) {
+    if (!isEvaluatorKind(typeValue) && !customAssertionTypes.has(typeValue)) {
       errors.push({
         severity: 'warning',
         filePath,
diff --git a/packages/core/src/evaluation/validation/file-type.ts b/packages/core/src/evaluation/validation/file-type.ts
index 46e87bef5..464610c61 100644
--- a/packages/core/src/evaluation/validation/file-type.ts
+++ b/packages/core/src/evaluation/validation/file-type.ts
@@ -20,6 +20,11 @@ export async function detectFileType(filePath: string): Promise<FileType> {
     const content = await readFile(filePath, 'utf8');
     const parsed = parse(content) as unknown;
 
+    // YAML array root → cases file (array of test case objects)
+    if (Array.isArray(parsed)) {
+      return 'cases';
+    }
+
     if (typeof parsed !== 'object' || parsed === null) {
       return inferFileTypeFromPath(filePath);
     }
@@ -65,8 +70,14 @@ function inferFileTypeFromPath(filePath: string): FileType {
     }
   }
 
-  // Default to eval file
-  return 'eval';
+  // Require .eval.yaml / .eval.yml suffix for eval files
+  const lower = basename.toLowerCase();
+  if (lower.endsWith('.eval.yaml') || lower.endsWith('.eval.yml')) {
+    return 'eval';
+  }
+
+  // Unrecognized — do not assume eval type
+  return 'unknown';
 }
 
 /**
diff --git a/packages/core/src/evaluation/validation/index.ts b/packages/core/src/evaluation/validation/index.ts
index 9347974f3..f5874ee9d 100644
--- a/packages/core/src/evaluation/validation/index.ts
+++ b/packages/core/src/evaluation/validation/index.ts
@@ -4,6 +4,7 @@
 
 export { detectFileType, isValidSchema, getExpectedSchema } from './file-type.js';
 export { validateEvalFile } from './eval-validator.js';
+export { validateCasesFile } from './cases-validator.js';
 export { validateTargetsFile } from './targets-validator.js';
 export { validateConfigFile } from './config-validator.js';
 export { validateFileReferences } from './file-reference-validator.js';
diff --git a/packages/core/src/evaluation/validation/types.ts b/packages/core/src/evaluation/validation/types.ts
index b49a0bed8..8aafe5aee 100644
--- a/packages/core/src/evaluation/validation/types.ts
+++ b/packages/core/src/evaluation/validation/types.ts
@@ -2,7 +2,7 @@
  * Validation result types for AgentV file validation.
  */
 
-export type FileType = 'eval' | 'targets' | 'config' | 'unknown';
+export type FileType = 'eval' | 'targets' | 'config' | 'cases' | 'unknown';
 
 export type ValidationSeverity = 'error' | 'warning';
 
diff --git a/packages/core/test/evaluation/validation/eval-validator.test.ts b/packages/core/test/evaluation/validation/eval-validator.test.ts
index f6b819479..7992160b8 100644
--- a/packages/core/test/evaluation/validation/eval-validator.test.ts
+++ b/packages/core/test/evaluation/validation/eval-validator.test.ts
@@ -993,7 +993,6 @@ tests:
         value: "world"
     metadata:
       tag: test
-    note: A note
 `,
       );
 
@@ -1051,8 +1050,8 @@ tests:
   });
 
   describe('removed legacy fields', () => {
-    it('warns on expected_outcome as unknown field', async () => {
-      const filePath = path.join(tempDir, 'expected-outcome-unknown.yaml');
+    it('warns on expected_outcome as deprecated field', async () => {
+      const filePath = path.join(tempDir, 'expected-outcome-deprecated.yaml');
       await writeFile(
         filePath,
         `tests:
@@ -1068,13 +1067,17 @@ tests:
 
       expect(result.valid).toBe(true);
       const warnings = result.errors.filter((e) => e.severity === 'warning');
-      expect(warnings.some((e) => e.message.includes("Unknown field 'expected_outcome'"))).toBe(
-        true,
-      );
+      expect(
+        warnings.some(
+          (e) =>
+            e.message.includes("'expected_outcome' is deprecated") &&
+            e.message.includes("'criteria'"),
+        ),
+      ).toBe(true);
     });
 
-    it('warns on assert as unknown field at test level', async () => {
-      const filePath = path.join(tempDir, 'assert-unknown.yaml');
+    it('warns on assert as deprecated field at test level', async () => {
+      const filePath = path.join(tempDir, 'assert-deprecated.yaml');
       await writeFile(
         filePath,
         `tests:
@@ -1090,11 +1093,15 @@ tests:
 
       expect(result.valid).toBe(true);
       const warnings = result.errors.filter((e) => e.severity === 'warning');
-      expect(warnings.some((e) => e.message.includes("Unknown field 'assert'"))).toBe(true);
+      expect(
+        warnings.some(
+          (e) => e.message.includes("'assert' is deprecated") && e.message.includes("'assertions'"),
+        ),
+      ).toBe(true);
     });
 
-    it('warns on assert as unknown field at top level', async () => {
-      const filePath = path.join(tempDir, 'assert-top-unknown.yaml');
+    it('warns on assert as deprecated field at top level', async () => {
+      const filePath = path.join(tempDir, 'assert-top-deprecated.yaml');
       await writeFile(
         filePath,
         `assert:
@@ -1109,7 +1116,11 @@ tests:
       const result = await validateEvalFile(filePath);
 
       const warnings = result.errors.filter((e) => e.severity === 'warning');
-      expect(warnings.some((e) => e.message.includes("Unknown field 'assert'"))).toBe(true);
+      expect(
+        warnings.some(
+          (e) => e.message.includes("'assert' is deprecated") && e.message.includes("'assertions'"),
+        ),
+      ).toBe(true);
     });
   });
 });
diff --git a/plugins/agentv-dev/skills/agentv-eval-writer/references/eval-schema.json b/plugins/agentv-dev/skills/agentv-eval-writer/references/eval-schema.json
index 80dc2ebd8..885d275b3 100644
--- a/plugins/agentv-dev/skills/agentv-eval-writer/references/eval-schema.json
+++ b/plugins/agentv-dev/skills/agentv-eval-writer/references/eval-schema.json
@@ -5102,9 +5102,6 @@
                   "suite": {
                     "type": "string"
                   },
-                  "note": {
-                    "type": "string"
-                  },
                   "depends_on": {
                     "type": "array",
                     "items": {
@@ -11373,9 +11370,6 @@
                   "suite": {
                     "type": "string"
                   },
-                  "note": {
-                    "type": "string"
-                  },
                   "depends_on": {
                     "type": "array",
                     "items": {