@wix/evalforge-types 0.16.0 → 0.18.0

package/build/index.js

@@ -26,6 +26,8 @@ __export(index_exports, {
   AllowedCommands: () => AllowedCommands,
   ApiCallSchema: () => ApiCallSchema,
   AssertionConfigSchema: () => AssertionConfigSchema,
+  AssertionParameterSchema: () => AssertionParameterSchema,
+  AssertionParameterTypeSchema: () => AssertionParameterTypeSchema,
   AssertionResultSchema: () => AssertionResultSchema,
   AssertionResultStatus: () => AssertionResultStatus,
   AssertionSchema: () => AssertionSchema,
@@ -92,6 +94,9 @@ __export(index_exports, {
   ProjectSchema: () => ProjectSchema,
   PromptResultSchema: () => PromptResultSchema,
   SKILL_FOLDER_NAME_REGEX: () => SKILL_FOLDER_NAME_REGEX,
+  SYSTEM_ASSERTIONS: () => SYSTEM_ASSERTIONS,
+  SYSTEM_ASSERTION_IDS: () => SYSTEM_ASSERTION_IDS,
+  ScenarioAssertionLinkSchema: () => ScenarioAssertionLinkSchema,
   SiteConfigTestSchema: () => SiteConfigTestSchema,
   SkillMetadataSchema: () => SkillMetadataSchema,
   SkillSchema: () => SkillSchema,
@@ -130,6 +135,9 @@ __export(index_exports, {
   getBuildPassedConfig: () => getBuildPassedConfig,
   getLlmJudgeConfig: () => getLlmJudgeConfig,
   getSkillWasCalledConfig: () => getSkillWasCalledConfig,
+  getSystemAssertion: () => getSystemAssertion,
+  getSystemAssertions: () => getSystemAssertions,
+  isSystemAssertionId: () => isSystemAssertionId,
   isValidSkillFolderName: () => isValidSkillFolderName,
   parseTraceEventLine: () => parseTraceEventLine,
   validateAssertionConfig: () => validateAssertionConfig
@@ -592,22 +600,145 @@ var EnvironmentSchema = import_zod19.z.object({
 });
 
 // src/scenario/test-scenario.ts
+var import_zod21 = require("zod");
+
+// src/assertion/assertion.ts
 var import_zod20 = require("zod");
-var ExpectedFileSchema = import_zod20.z.object({
+var AssertionTypeSchema = import_zod20.z.enum([
+  "skill_was_called",
+  "build_passed",
+  "llm_judge"
+]);
+var AssertionParameterTypeSchema = import_zod20.z.enum([
+  "string",
+  "number",
+  "boolean"
+]);
+var AssertionParameterSchema = import_zod20.z.object({
+  /** Parameter name (used as key in params object) */
+  name: import_zod20.z.string().min(1),
+  /** Display label for the parameter */
+  label: import_zod20.z.string().min(1),
+  /** Parameter type */
+  type: AssertionParameterTypeSchema,
+  /** Whether this parameter is required */
+  required: import_zod20.z.boolean(),
+  /** Default value (optional, used when not provided) */
+  defaultValue: import_zod20.z.union([import_zod20.z.string(), import_zod20.z.number(), import_zod20.z.boolean()]).optional()
+});
+var ScenarioAssertionLinkSchema = import_zod20.z.object({
+  /** ID of the assertion (can be system assertion like 'system:skill_was_called' or custom assertion UUID) */
+  assertionId: import_zod20.z.string(),
+  /** Parameter values for this assertion in this scenario */
+  params: import_zod20.z.record(
+    import_zod20.z.string(),
+    import_zod20.z.union([import_zod20.z.string(), import_zod20.z.number(), import_zod20.z.boolean(), import_zod20.z.null()])
+  ).optional()
+});
+var SkillWasCalledConfigSchema = import_zod20.z.object({
+  /** Name of the skill that must have been called */
+  skillName: import_zod20.z.string().min(1)
+});
+var BuildPassedConfigSchema = import_zod20.z.strictObject({
+  /** Command to run (default: "yarn build") */
+  command: import_zod20.z.string().optional(),
+  /** Expected exit code (default: 0) */
+  expectedExitCode: import_zod20.z.number().int().optional()
+});
+var LlmJudgeConfigSchema = import_zod20.z.object({
+  /**
+   * Prompt template with placeholders:
+   * - {{output}}: agent's final output
+   * - {{cwd}}: working directory
+   * - {{changedFiles}}: all files changed (new, modified)
+   * - {{modifiedFiles}}: only existing files that were modified
+   * - {{newFiles}}: only new files that were created
+   * - {{trace}}: step-by-step trace of tool calls
+   * - Custom parameters defined in the parameters array
+   */
+  prompt: import_zod20.z.string().min(1),
+  /** Optional system prompt for the judge */
+  systemPrompt: import_zod20.z.string().optional(),
+  /** Minimum score to pass (0-100, default 70) */
+  minScore: import_zod20.z.number().int().min(0).max(100).optional(),
+  /** Model for the judge (e.g. claude-3-5-haiku-20241022) */
+  model: import_zod20.z.string().optional(),
+  /** Max output tokens */
+  maxTokens: import_zod20.z.number().int().optional(),
+  /** Temperature (0-1) */
+  temperature: import_zod20.z.number().min(0).max(1).optional(),
+  /** User-defined parameters for this assertion */
+  parameters: import_zod20.z.array(AssertionParameterSchema).optional()
+});
+var AssertionConfigSchema = import_zod20.z.union([
+  LlmJudgeConfigSchema,
+  // requires prompt - check first
+  SkillWasCalledConfigSchema,
+  // requires skillName
+  BuildPassedConfigSchema,
+  // all optional, uses strictObject to reject unknown keys
+  import_zod20.z.object({})
+  // fallback empty config
+]);
+var CustomAssertionSchema = TenantEntitySchema.extend({
+  /** The assertion type */
+  type: AssertionTypeSchema,
+  /** Type-specific configuration */
+  config: AssertionConfigSchema
+});
+var CreateCustomAssertionInputSchema = CustomAssertionSchema.omit({
+  id: true,
+  createdAt: true,
+  updatedAt: true,
+  deleted: true
+});
+var UpdateCustomAssertionInputSchema = CreateCustomAssertionInputSchema.partial();
+function validateAssertionConfig(type, config) {
+  switch (type) {
+    case "skill_was_called":
+      return SkillWasCalledConfigSchema.safeParse(config).success;
+    case "build_passed":
+      return BuildPassedConfigSchema.safeParse(config).success;
+    case "llm_judge":
+      return LlmJudgeConfigSchema.safeParse(config).success;
+    default:
+      return false;
+  }
+}
+function getSkillWasCalledConfig(assertion) {
+  if (assertion.type !== "skill_was_called") return null;
+  const result = SkillWasCalledConfigSchema.safeParse(assertion.config);
+  return result.success ? result.data : null;
+}
+function getBuildPassedConfig(assertion) {
+  if (assertion.type !== "build_passed") return null;
+  const result = BuildPassedConfigSchema.safeParse(assertion.config);
+  return result.success ? result.data : null;
+}
+function getLlmJudgeConfig(assertion) {
+  if (assertion.type !== "llm_judge") return null;
+  const result = LlmJudgeConfigSchema.safeParse(assertion.config);
+  return result.success ? result.data : null;
+}
+
+// src/scenario/test-scenario.ts
+var ExpectedFileSchema = import_zod21.z.object({
   /** Relative path where the file should be created */
-  path: import_zod20.z.string(),
+  path: import_zod21.z.string(),
   /** Optional expected content */
-  content: import_zod20.z.string().optional()
+  content: import_zod21.z.string().optional()
 });
 var TestScenarioSchema = TenantEntitySchema.extend({
   /** The prompt sent to the agent to trigger the task */
-  triggerPrompt: import_zod20.z.string().min(10),
+  triggerPrompt: import_zod21.z.string().min(10),
   /** ID of the template to use for this scenario (null = no template) */
-  templateId: import_zod20.z.string().nullish(),
+  templateId: import_zod21.z.string().nullish(),
   /** Inline assertions to evaluate for this scenario (legacy) */
-  assertions: import_zod20.z.array(AssertionSchema).optional(),
-  /** IDs of saved assertions to evaluate (from assertions table) */
-  assertionIds: import_zod20.z.array(import_zod20.z.string()).optional()
+  assertions: import_zod21.z.array(AssertionSchema).optional(),
+  /** IDs of saved assertions to evaluate (from assertions table) - legacy, use assertionLinks */
+  assertionIds: import_zod21.z.array(import_zod21.z.string()).optional(),
+  /** Linked assertions with per-scenario parameter values */
+  assertionLinks: import_zod21.z.array(ScenarioAssertionLinkSchema).optional()
 });
 var CreateTestScenarioInputSchema = TestScenarioSchema.omit({
   id: true,
@@ -618,10 +749,10 @@ var CreateTestScenarioInputSchema = TestScenarioSchema.omit({
 var UpdateTestScenarioInputSchema = CreateTestScenarioInputSchema.partial();
 
 // src/suite/test-suite.ts
-var import_zod21 = require("zod");
+var import_zod22 = require("zod");
 var TestSuiteSchema = TenantEntitySchema.extend({
   /** IDs of test scenarios in this suite */
-  scenarioIds: import_zod21.z.array(import_zod21.z.string())
+  scenarioIds: import_zod22.z.array(import_zod22.z.string())
 });
 var CreateTestSuiteInputSchema = TestSuiteSchema.omit({
   id: true,
@@ -632,21 +763,21 @@ var CreateTestSuiteInputSchema = TestSuiteSchema.omit({
 var UpdateTestSuiteInputSchema = CreateTestSuiteInputSchema.partial();
 
 // src/evaluation/metrics.ts
-var import_zod22 = require("zod");
-var TokenUsageSchema = import_zod22.z.object({
-  prompt: import_zod22.z.number(),
-  completion: import_zod22.z.number(),
-  total: import_zod22.z.number()
-});
-var EvalMetricsSchema = import_zod22.z.object({
-  totalAssertions: import_zod22.z.number(),
-  passed: import_zod22.z.number(),
-  failed: import_zod22.z.number(),
-  skipped: import_zod22.z.number(),
-  errors: import_zod22.z.number(),
-  passRate: import_zod22.z.number(),
-  avgDuration: import_zod22.z.number(),
-  totalDuration: import_zod22.z.number()
+var import_zod23 = require("zod");
+var TokenUsageSchema = import_zod23.z.object({
+  prompt: import_zod23.z.number(),
+  completion: import_zod23.z.number(),
+  total: import_zod23.z.number()
+});
+var EvalMetricsSchema = import_zod23.z.object({
+  totalAssertions: import_zod23.z.number(),
+  passed: import_zod23.z.number(),
+  failed: import_zod23.z.number(),
+  skipped: import_zod23.z.number(),
+  errors: import_zod23.z.number(),
+  passRate: import_zod23.z.number(),
+  avgDuration: import_zod23.z.number(),
+  totalDuration: import_zod23.z.number()
 });
 var EvalStatus = /* @__PURE__ */ ((EvalStatus2) => {
   EvalStatus2["PENDING"] = "pending";
@@ -656,7 +787,7 @@ var EvalStatus = /* @__PURE__ */ ((EvalStatus2) => {
   EvalStatus2["CANCELLED"] = "cancelled";
   return EvalStatus2;
 })(EvalStatus || {});
-var EvalStatusSchema = import_zod22.z.enum(EvalStatus);
+var EvalStatusSchema = import_zod23.z.enum(EvalStatus);
 var LLMStepType = /* @__PURE__ */ ((LLMStepType2) => {
   LLMStepType2["COMPLETION"] = "completion";
   LLMStepType2["TOOL_USE"] = "tool_use";
@@ -664,52 +795,52 @@ var LLMStepType = /* @__PURE__ */ ((LLMStepType2) => {
   LLMStepType2["THINKING"] = "thinking";
   return LLMStepType2;
 })(LLMStepType || {});
-var LLMTraceStepSchema = import_zod22.z.object({
-  id: import_zod22.z.string(),
-  stepNumber: import_zod22.z.number(),
-  type: import_zod22.z.enum(LLMStepType),
-  model: import_zod22.z.string(),
-  provider: import_zod22.z.string(),
-  startedAt: import_zod22.z.string(),
-  durationMs: import_zod22.z.number(),
+var LLMTraceStepSchema = import_zod23.z.object({
+  id: import_zod23.z.string(),
+  stepNumber: import_zod23.z.number(),
+  type: import_zod23.z.enum(LLMStepType),
+  model: import_zod23.z.string(),
+  provider: import_zod23.z.string(),
+  startedAt: import_zod23.z.string(),
+  durationMs: import_zod23.z.number(),
   tokenUsage: TokenUsageSchema,
-  costUsd: import_zod22.z.number(),
-  toolName: import_zod22.z.string().optional(),
-  toolArguments: import_zod22.z.string().optional(),
-  inputPreview: import_zod22.z.string().optional(),
-  outputPreview: import_zod22.z.string().optional(),
-  success: import_zod22.z.boolean(),
-  error: import_zod22.z.string().optional()
-});
-var LLMBreakdownStatsSchema = import_zod22.z.object({
-  count: import_zod22.z.number(),
-  durationMs: import_zod22.z.number(),
-  tokens: import_zod22.z.number(),
-  costUsd: import_zod22.z.number()
-});
-var LLMTraceSummarySchema = import_zod22.z.object({
-  totalSteps: import_zod22.z.number(),
-  totalDurationMs: import_zod22.z.number(),
+  costUsd: import_zod23.z.number(),
+  toolName: import_zod23.z.string().optional(),
+  toolArguments: import_zod23.z.string().optional(),
+  inputPreview: import_zod23.z.string().optional(),
+  outputPreview: import_zod23.z.string().optional(),
+  success: import_zod23.z.boolean(),
+  error: import_zod23.z.string().optional()
+});
+var LLMBreakdownStatsSchema = import_zod23.z.object({
+  count: import_zod23.z.number(),
+  durationMs: import_zod23.z.number(),
+  tokens: import_zod23.z.number(),
+  costUsd: import_zod23.z.number()
+});
+var LLMTraceSummarySchema = import_zod23.z.object({
+  totalSteps: import_zod23.z.number(),
+  totalDurationMs: import_zod23.z.number(),
   totalTokens: TokenUsageSchema,
-  totalCostUsd: import_zod22.z.number(),
-  stepTypeBreakdown: import_zod22.z.record(import_zod22.z.string(), LLMBreakdownStatsSchema).optional(),
-  modelBreakdown: import_zod22.z.record(import_zod22.z.string(), LLMBreakdownStatsSchema),
-  modelsUsed: import_zod22.z.array(import_zod22.z.string())
-});
-var LLMTraceSchema = import_zod22.z.object({
-  id: import_zod22.z.string(),
-  steps: import_zod22.z.array(LLMTraceStepSchema),
+  totalCostUsd: import_zod23.z.number(),
+  stepTypeBreakdown: import_zod23.z.record(import_zod23.z.string(), LLMBreakdownStatsSchema).optional(),
+  modelBreakdown: import_zod23.z.record(import_zod23.z.string(), LLMBreakdownStatsSchema),
+  modelsUsed: import_zod23.z.array(import_zod23.z.string())
+});
+var LLMTraceSchema = import_zod23.z.object({
+  id: import_zod23.z.string(),
+  steps: import_zod23.z.array(LLMTraceStepSchema),
   summary: LLMTraceSummarySchema
 });
 
 // src/evaluation/eval-result.ts
-var import_zod25 = require("zod");
+var import_zod26 = require("zod");
 
 // src/evaluation/eval-run.ts
-var import_zod24 = require("zod");
+var import_zod25 = require("zod");
 
 // src/evaluation/live-trace.ts
-var import_zod23 = require("zod");
+var import_zod24 = require("zod");
 var LiveTraceEventType = /* @__PURE__ */ ((LiveTraceEventType2) => {
   LiveTraceEventType2["THINKING"] = "thinking";
   LiveTraceEventType2["TOOL_USE"] = "tool_use";
@@ -723,37 +854,37 @@ var LiveTraceEventType = /* @__PURE__ */ ((LiveTraceEventType2) => {
   LiveTraceEventType2["USER"] = "user";
   return LiveTraceEventType2;
 })(LiveTraceEventType || {});
-var LiveTraceEventSchema = import_zod23.z.object({
+var LiveTraceEventSchema = import_zod24.z.object({
   /** The evaluation run ID */
-  evalRunId: import_zod23.z.string(),
+  evalRunId: import_zod24.z.string(),
   /** The scenario ID being executed */
-  scenarioId: import_zod23.z.string(),
+  scenarioId: import_zod24.z.string(),
   /** The scenario name for display */
-  scenarioName: import_zod23.z.string(),
+  scenarioName: import_zod24.z.string(),
   /** The target ID (skill, agent, etc.) */
-  targetId: import_zod23.z.string(),
+  targetId: import_zod24.z.string(),
   /** The target name for display */
-  targetName: import_zod23.z.string(),
+  targetName: import_zod24.z.string(),
   /** Step number in the current scenario execution */
-  stepNumber: import_zod23.z.number(),
+  stepNumber: import_zod24.z.number(),
   /** Type of trace event */
-  type: import_zod23.z.enum(LiveTraceEventType),
+  type: import_zod24.z.enum(LiveTraceEventType),
   /** Tool name if this is a tool_use event */
-  toolName: import_zod23.z.string().optional(),
+  toolName: import_zod24.z.string().optional(),
   /** Tool arguments preview (truncated JSON) */
-  toolArgs: import_zod23.z.string().optional(),
+  toolArgs: import_zod24.z.string().optional(),
   /** Output preview (truncated text) */
-  outputPreview: import_zod23.z.string().optional(),
+  outputPreview: import_zod24.z.string().optional(),
   /** File path for file operations */
-  filePath: import_zod23.z.string().optional(),
+  filePath: import_zod24.z.string().optional(),
   /** Elapsed time in milliseconds for progress events */
-  elapsedMs: import_zod23.z.number().optional(),
+  elapsedMs: import_zod24.z.number().optional(),
   /** Thinking/reasoning text from Claude */
-  thinking: import_zod23.z.string().optional(),
+  thinking: import_zod24.z.string().optional(),
   /** Timestamp when this event occurred */
-  timestamp: import_zod23.z.string(),
+  timestamp: import_zod24.z.string(),
   /** Whether this is the final event for this scenario */
-  isComplete: import_zod23.z.boolean()
+  isComplete: import_zod24.z.boolean()
 });
 var TRACE_EVENT_PREFIX = "TRACE_EVENT:";
 function parseTraceEventLine(line) {
@@ -781,14 +912,14 @@ var TriggerType = /* @__PURE__ */ ((TriggerType2) => {
   TriggerType2["MANUAL"] = "MANUAL";
   return TriggerType2;
 })(TriggerType || {});
-var TriggerMetadataSchema = import_zod24.z.object({
-  version: import_zod24.z.string().optional(),
-  resourceUpdated: import_zod24.z.array(import_zod24.z.string()).optional()
+var TriggerMetadataSchema = import_zod25.z.object({
+  version: import_zod25.z.string().optional(),
+  resourceUpdated: import_zod25.z.array(import_zod25.z.string()).optional()
 });
-var TriggerSchema = import_zod24.z.object({
-  id: import_zod24.z.string(),
+var TriggerSchema = import_zod25.z.object({
+  id: import_zod25.z.string(),
   metadata: TriggerMetadataSchema.optional(),
-  type: import_zod24.z.enum(TriggerType)
+  type: import_zod25.z.enum(TriggerType)
 });
 var FailureCategory = /* @__PURE__ */ ((FailureCategory2) => {
   FailureCategory2["MISSING_FILE"] = "missing_file";
@@ -806,28 +937,28 @@ var FailureSeverity = /* @__PURE__ */ ((FailureSeverity2) => {
   FailureSeverity2["LOW"] = "low";
   return FailureSeverity2;
 })(FailureSeverity || {});
-var DiffLineTypeSchema = import_zod24.z.enum(["added", "removed", "unchanged"]);
-var DiffLineSchema = import_zod24.z.object({
+var DiffLineTypeSchema = import_zod25.z.enum(["added", "removed", "unchanged"]);
+var DiffLineSchema = import_zod25.z.object({
   type: DiffLineTypeSchema,
-  content: import_zod24.z.string(),
-  lineNumber: import_zod24.z.number()
-});
-var DiffContentSchema = import_zod24.z.object({
-  path: import_zod24.z.string(),
-  expected: import_zod24.z.string(),
-  actual: import_zod24.z.string(),
-  diffLines: import_zod24.z.array(DiffLineSchema),
-  renamedFrom: import_zod24.z.string().optional()
-});
-var CommandExecutionSchema = import_zod24.z.object({
-  command: import_zod24.z.string(),
-  exitCode: import_zod24.z.number(),
-  output: import_zod24.z.string().optional(),
-  duration: import_zod24.z.number()
-});
-var FileModificationSchema = import_zod24.z.object({
-  path: import_zod24.z.string(),
-  action: import_zod24.z.enum(["created", "modified", "deleted"])
+  content: import_zod25.z.string(),
+  lineNumber: import_zod25.z.number()
+});
+var DiffContentSchema = import_zod25.z.object({
+  path: import_zod25.z.string(),
+  expected: import_zod25.z.string(),
+  actual: import_zod25.z.string(),
+  diffLines: import_zod25.z.array(DiffLineSchema),
+  renamedFrom: import_zod25.z.string().optional()
+});
+var CommandExecutionSchema = import_zod25.z.object({
+  command: import_zod25.z.string(),
+  exitCode: import_zod25.z.number(),
+  output: import_zod25.z.string().optional(),
+  duration: import_zod25.z.number()
+});
+var FileModificationSchema = import_zod25.z.object({
+  path: import_zod25.z.string(),
+  action: import_zod25.z.enum(["created", "modified", "deleted"])
 });
 var TemplateFileStatus = /* @__PURE__ */ ((TemplateFileStatus2) => {
   TemplateFileStatus2["NEW"] = "new";
@@ -835,75 +966,75 @@ var TemplateFileStatus = /* @__PURE__ */ ((TemplateFileStatus2) => {
   TemplateFileStatus2["UNCHANGED"] = "unchanged";
   return TemplateFileStatus2;
 })(TemplateFileStatus || {});
-var TemplateFileSchema = import_zod24.z.object({
+var TemplateFileSchema = import_zod25.z.object({
   /** Relative path within the template */
-  path: import_zod24.z.string(),
+  path: import_zod25.z.string(),
   /** Full file content after execution */
-  content: import_zod24.z.string(),
+  content: import_zod25.z.string(),
   /** File status (new, modified, unchanged) */
-  status: import_zod24.z.enum(["new", "modified", "unchanged"])
-});
-var ApiCallSchema = import_zod24.z.object({
-  endpoint: import_zod24.z.string(),
-  tokensUsed: import_zod24.z.number(),
-  duration: import_zod24.z.number()
-});
-var ExecutionTraceSchema = import_zod24.z.object({
-  commands: import_zod24.z.array(CommandExecutionSchema),
-  filesModified: import_zod24.z.array(FileModificationSchema),
-  apiCalls: import_zod24.z.array(ApiCallSchema),
-  totalDuration: import_zod24.z.number()
-});
-var FailureAnalysisSchema = import_zod24.z.object({
-  category: import_zod24.z.enum(FailureCategory),
-  severity: import_zod24.z.enum(FailureSeverity),
-  summary: import_zod24.z.string(),
-  details: import_zod24.z.string(),
-  rootCause: import_zod24.z.string(),
-  suggestedFix: import_zod24.z.string(),
-  relatedAssertions: import_zod24.z.array(import_zod24.z.string()),
-  codeSnippet: import_zod24.z.string().optional(),
-  similarIssues: import_zod24.z.array(import_zod24.z.string()).optional(),
-  patternId: import_zod24.z.string().optional(),
+  status: import_zod25.z.enum(["new", "modified", "unchanged"])
+});
+var ApiCallSchema = import_zod25.z.object({
+  endpoint: import_zod25.z.string(),
+  tokensUsed: import_zod25.z.number(),
+  duration: import_zod25.z.number()
+});
+var ExecutionTraceSchema = import_zod25.z.object({
+  commands: import_zod25.z.array(CommandExecutionSchema),
+  filesModified: import_zod25.z.array(FileModificationSchema),
+  apiCalls: import_zod25.z.array(ApiCallSchema),
+  totalDuration: import_zod25.z.number()
+});
+var FailureAnalysisSchema = import_zod25.z.object({
+  category: import_zod25.z.enum(FailureCategory),
+  severity: import_zod25.z.enum(FailureSeverity),
+  summary: import_zod25.z.string(),
+  details: import_zod25.z.string(),
+  rootCause: import_zod25.z.string(),
+  suggestedFix: import_zod25.z.string(),
+  relatedAssertions: import_zod25.z.array(import_zod25.z.string()),
+  codeSnippet: import_zod25.z.string().optional(),
+  similarIssues: import_zod25.z.array(import_zod25.z.string()).optional(),
+  patternId: import_zod25.z.string().optional(),
   // Extended fields for detailed debugging
   diff: DiffContentSchema.optional(),
   executionTrace: ExecutionTraceSchema.optional()
 });
 var EvalRunSchema = TenantEntitySchema.extend({
   /** Agent ID for this run */
-  agentId: import_zod24.z.string().optional(),
+  agentId: import_zod25.z.string().optional(),
   /** Skills group ID for this run */
-  skillsGroupId: import_zod24.z.string().optional(),
+  skillsGroupId: import_zod25.z.string().optional(),
   /** Scenario IDs to run */
-  scenarioIds: import_zod24.z.array(import_zod24.z.string()),
+  scenarioIds: import_zod25.z.array(import_zod25.z.string()),
   /** Current status */
   status: EvalStatusSchema,
   /** Progress percentage (0-100) */
-  progress: import_zod24.z.number(),
+  progress: import_zod25.z.number(),
   /** Results for each scenario/target combination (lazy to break eval-result ↔ eval-run cycle) */
-  results: import_zod24.z.array(import_zod24.z.lazy(() => EvalRunResultSchema)),
+  results: import_zod25.z.array(import_zod25.z.lazy(() => EvalRunResultSchema)),
   /** Aggregated metrics across all results */
   aggregateMetrics: EvalMetricsSchema,
   /** Failure analyses */
-  failureAnalyses: import_zod24.z.array(FailureAnalysisSchema).optional(),
+  failureAnalyses: import_zod25.z.array(FailureAnalysisSchema).optional(),
   /** Aggregated LLM trace summary */
   llmTraceSummary: LLMTraceSummarySchema.optional(),
   /** What triggered this run */
   trigger: TriggerSchema.optional(),
   /** When the run started (set when evaluation is triggered) */
-  startedAt: import_zod24.z.string().optional(),
+  startedAt: import_zod25.z.string().optional(),
   /** When the run completed */
-  completedAt: import_zod24.z.string().optional(),
+  completedAt: import_zod25.z.string().optional(),
   /** Live trace events captured during execution (for playback on results page) */
-  liveTraceEvents: import_zod24.z.array(LiveTraceEventSchema).optional(),
+  liveTraceEvents: import_zod25.z.array(LiveTraceEventSchema).optional(),
   /** Remote job ID for tracking execution in Dev Machines */
-  jobId: import_zod24.z.string().optional(),
+  jobId: import_zod25.z.string().optional(),
   /** Remote job status from the Dev Machine API (PENDING, RUNNING, COMPLETED, FAILED, CANCELLED) */
-  jobStatus: import_zod24.z.string().optional(),
+  jobStatus: import_zod25.z.string().optional(),
   /** Remote job error message if the job failed */
-  jobError: import_zod24.z.string().optional(),
+  jobError: import_zod25.z.string().optional(),
   /** Timestamp of the last job status check */
-  jobStatusCheckedAt: import_zod24.z.string().optional()
+  jobStatusCheckedAt: import_zod25.z.string().optional()
 });
 var CreateEvalRunInputSchema = EvalRunSchema.omit({
   id: true,
@@ -916,28 +1047,28 @@ var CreateEvalRunInputSchema = EvalRunSchema.omit({
   startedAt: true,
   completedAt: true
 });
-var EvaluationProgressSchema = import_zod24.z.object({
-  runId: import_zod24.z.string(),
-  targetId: import_zod24.z.string(),
-  totalScenarios: import_zod24.z.number(),
-  completedScenarios: import_zod24.z.number(),
-  scenarioProgress: import_zod24.z.array(
-    import_zod24.z.object({
-      scenarioId: import_zod24.z.string(),
-      currentStep: import_zod24.z.string(),
-      error: import_zod24.z.string().optional()
+var EvaluationProgressSchema = import_zod25.z.object({
+  runId: import_zod25.z.string(),
+  targetId: import_zod25.z.string(),
+  totalScenarios: import_zod25.z.number(),
+  completedScenarios: import_zod25.z.number(),
+  scenarioProgress: import_zod25.z.array(
+    import_zod25.z.object({
+      scenarioId: import_zod25.z.string(),
+      currentStep: import_zod25.z.string(),
+      error: import_zod25.z.string().optional()
     })
   ),
-  createdAt: import_zod24.z.number()
+  createdAt: import_zod25.z.number()
 });
-var EvaluationLogSchema = import_zod24.z.object({
-  runId: import_zod24.z.string(),
-  scenarioId: import_zod24.z.string(),
-  log: import_zod24.z.object({
-    level: import_zod24.z.enum(["info", "error", "debug"]),
-    message: import_zod24.z.string().optional(),
-    args: import_zod24.z.array(import_zod24.z.any()).optional(),
-    error: import_zod24.z.string().optional()
+var EvaluationLogSchema = import_zod25.z.object({
+  runId: import_zod25.z.string(),
+  scenarioId: import_zod25.z.string(),
+  log: import_zod25.z.object({
+    level: import_zod25.z.enum(["info", "error", "debug"]),
+    message: import_zod25.z.string().optional(),
+    args: import_zod25.z.array(import_zod25.z.any()).optional(),
+    error: import_zod25.z.string().optional()
   })
 });
 var LLM_TIMEOUT = 12e4;
@@ -950,91 +1081,91 @@ var AssertionResultStatus = /* @__PURE__ */ ((AssertionResultStatus2) => {
   AssertionResultStatus2["ERROR"] = "error";
   return AssertionResultStatus2;
 })(AssertionResultStatus || {});
-var AssertionResultSchema = import_zod25.z.object({
-  id: import_zod25.z.string(),
-  assertionId: import_zod25.z.string(),
-  assertionType: import_zod25.z.string(),
-  assertionName: import_zod25.z.string(),
-  status: import_zod25.z.enum(AssertionResultStatus),
-  message: import_zod25.z.string().optional(),
-  expected: import_zod25.z.string().optional(),
-  actual: import_zod25.z.string().optional(),
-  duration: import_zod25.z.number().optional(),
-  details: import_zod25.z.record(import_zod25.z.string(), import_zod25.z.unknown()).optional(),
-  llmTraceSteps: import_zod25.z.array(LLMTraceStepSchema).optional()
-});
-var EvalRunResultSchema = import_zod25.z.object({
-  id: import_zod25.z.string(),
-  targetId: import_zod25.z.string(),
-  targetName: import_zod25.z.string().optional(),
-  scenarioId: import_zod25.z.string(),
-  scenarioName: import_zod25.z.string(),
+var AssertionResultSchema = import_zod26.z.object({
+  id: import_zod26.z.string(),
+  assertionId: import_zod26.z.string(),
+  assertionType: import_zod26.z.string(),
+  assertionName: import_zod26.z.string(),
+  status: import_zod26.z.enum(AssertionResultStatus),
+  message: import_zod26.z.string().optional(),
+  expected: import_zod26.z.string().optional(),
+  actual: import_zod26.z.string().optional(),
+  duration: import_zod26.z.number().optional(),
+  details: import_zod26.z.record(import_zod26.z.string(), import_zod26.z.unknown()).optional(),
+  llmTraceSteps: import_zod26.z.array(LLMTraceStepSchema).optional()
+});
+var EvalRunResultSchema = import_zod26.z.object({
+  id: import_zod26.z.string(),
+  targetId: import_zod26.z.string(),
+  targetName: import_zod26.z.string().optional(),
+  scenarioId: import_zod26.z.string(),
+  scenarioName: import_zod26.z.string(),
   modelConfig: ModelConfigSchema.optional(),
-  assertionResults: import_zod25.z.array(AssertionResultSchema),
+  assertionResults: import_zod26.z.array(AssertionResultSchema),
   metrics: EvalMetricsSchema.optional(),
-  passed: import_zod25.z.number(),
-  failed: import_zod25.z.number(),
-  passRate: import_zod25.z.number(),
-  duration: import_zod25.z.number(),
-  outputText: import_zod25.z.string().optional(),
-  files: import_zod25.z.array(ExpectedFileSchema).optional(),
-  fileDiffs: import_zod25.z.array(DiffContentSchema).optional(),
+  passed: import_zod26.z.number(),
+  failed: import_zod26.z.number(),
+  passRate: import_zod26.z.number(),
+  duration: import_zod26.z.number(),
+  outputText: import_zod26.z.string().optional(),
+  files: import_zod26.z.array(ExpectedFileSchema).optional(),
+  fileDiffs: import_zod26.z.array(DiffContentSchema).optional(),
   /** Full template files after execution with status indicators */
-  templateFiles: import_zod25.z.array(TemplateFileSchema).optional(),
-  startedAt: import_zod25.z.string().optional(),
-  completedAt: import_zod25.z.string().optional(),
+  templateFiles: import_zod26.z.array(TemplateFileSchema).optional(),
+  startedAt: import_zod26.z.string().optional(),
+  completedAt: import_zod26.z.string().optional(),
   llmTrace: LLMTraceSchema.optional()
 });
-var PromptResultSchema = import_zod25.z.object({
-  text: import_zod25.z.string(),
-  files: import_zod25.z.array(import_zod25.z.unknown()).optional(),
-  finishReason: import_zod25.z.string().optional(),
-  reasoning: import_zod25.z.string().optional(),
-  reasoningDetails: import_zod25.z.unknown().optional(),
-  toolCalls: import_zod25.z.array(import_zod25.z.unknown()).optional(),
-  toolResults: import_zod25.z.array(import_zod25.z.unknown()).optional(),
-  warnings: import_zod25.z.array(import_zod25.z.unknown()).optional(),
-  sources: import_zod25.z.array(import_zod25.z.unknown()).optional(),
-  steps: import_zod25.z.array(import_zod25.z.unknown()),
-  generationTimeMs: import_zod25.z.number(),
-  prompt: import_zod25.z.string(),
-  systemPrompt: import_zod25.z.string(),
-  usage: import_zod25.z.object({
-    totalTokens: import_zod25.z.number().optional(),
-    totalMicrocentsSpent: import_zod25.z.number().optional()
+var PromptResultSchema = import_zod26.z.object({
+  text: import_zod26.z.string(),
+  files: import_zod26.z.array(import_zod26.z.unknown()).optional(),
+  finishReason: import_zod26.z.string().optional(),
+  reasoning: import_zod26.z.string().optional(),
+  reasoningDetails: import_zod26.z.unknown().optional(),
+  toolCalls: import_zod26.z.array(import_zod26.z.unknown()).optional(),
+  toolResults: import_zod26.z.array(import_zod26.z.unknown()).optional(),
+  warnings: import_zod26.z.array(import_zod26.z.unknown()).optional(),
+  sources: import_zod26.z.array(import_zod26.z.unknown()).optional(),
+  steps: import_zod26.z.array(import_zod26.z.unknown()),
+  generationTimeMs: import_zod26.z.number(),
+  prompt: import_zod26.z.string(),
+  systemPrompt: import_zod26.z.string(),
+  usage: import_zod26.z.object({
+    totalTokens: import_zod26.z.number().optional(),
+    totalMicrocentsSpent: import_zod26.z.number().optional()
   })
 });
-var EvaluationResultSchema = import_zod25.z.object({
-  id: import_zod25.z.string(),
-  runId: import_zod25.z.string(),
-  timestamp: import_zod25.z.number(),
+var EvaluationResultSchema = import_zod26.z.object({
+  id: import_zod26.z.string(),
+  runId: import_zod26.z.string(),
+  timestamp: import_zod26.z.number(),
   promptResult: PromptResultSchema,
-  testResults: import_zod25.z.array(import_zod25.z.unknown()),
-  tags: import_zod25.z.array(import_zod25.z.string()).optional(),
-  feedback: import_zod25.z.string().optional(),
-  score: import_zod25.z.number(),
-  suiteId: import_zod25.z.string().optional()
-});
-var LeanEvaluationResultSchema = import_zod25.z.object({
-  id: import_zod25.z.string(),
-  runId: import_zod25.z.string(),
-  timestamp: import_zod25.z.number(),
-  tags: import_zod25.z.array(import_zod25.z.string()).optional(),
-  scenarioId: import_zod25.z.string(),
-  scenarioVersion: import_zod25.z.number().optional(),
-  targetId: import_zod25.z.string(),
-  targetVersion: import_zod25.z.number().optional(),
-  suiteId: import_zod25.z.string().optional(),
-  score: import_zod25.z.number(),
-  time: import_zod25.z.number().optional(),
-  microcentsSpent: import_zod25.z.number().optional()
+  testResults: import_zod26.z.array(import_zod26.z.unknown()),
+  tags: import_zod26.z.array(import_zod26.z.string()).optional(),
+  feedback: import_zod26.z.string().optional(),
+  score: import_zod26.z.number(),
+  suiteId: import_zod26.z.string().optional()
+});
+var LeanEvaluationResultSchema = import_zod26.z.object({
+  id: import_zod26.z.string(),
+  runId: import_zod26.z.string(),
+  timestamp: import_zod26.z.number(),
+  tags: import_zod26.z.array(import_zod26.z.string()).optional(),
+  scenarioId: import_zod26.z.string(),
+  scenarioVersion: import_zod26.z.number().optional(),
+  targetId: import_zod26.z.string(),
+  targetVersion: import_zod26.z.number().optional(),
+  suiteId: import_zod26.z.string().optional(),
+  score: import_zod26.z.number(),
+  time: import_zod26.z.number().optional(),
+  microcentsSpent: import_zod26.z.number().optional()
 });
 
 // src/project/project.ts
-var import_zod26 = require("zod");
+var import_zod27 = require("zod");
 var ProjectSchema = BaseEntitySchema.extend({
-  appId: import_zod26.z.string().optional().describe("The ID of the app in Dev Center"),
-  appSecret: import_zod26.z.string().optional().describe("The secret of the app in Dev Center")
+  appId: import_zod27.z.string().optional().describe("The ID of the app in Dev Center"),
+  appSecret: import_zod27.z.string().optional().describe("The secret of the app in Dev Center")
 });
 var CreateProjectInputSchema = ProjectSchema.omit({
   id: true,
@@ -1045,10 +1176,10 @@ var CreateProjectInputSchema = ProjectSchema.omit({
 var UpdateProjectInputSchema = CreateProjectInputSchema.partial();
 
 // src/template/template.ts
-var import_zod27 = require("zod");
+var import_zod28 = require("zod");
 var TemplateSchema = TenantEntitySchema.extend({
   /** URL to download the template from */
-  downloadUrl: import_zod27.z.url()
+  downloadUrl: import_zod28.z.url()
 });
 var CreateTemplateInputSchema = TemplateSchema.omit({
   id: true,
@@ -1058,86 +1189,69 @@ var CreateTemplateInputSchema = TemplateSchema.omit({
 });
 var UpdateTemplateInputSchema = CreateTemplateInputSchema.partial();
 
-// src/assertion/assertion.ts
-var import_zod28 = require("zod");
-var AssertionTypeSchema = import_zod28.z.enum([
-  "skill_was_called",
-  "build_passed",
-  "llm_judge",
-  "custom"
-]);
-var SkillWasCalledConfigSchema = import_zod28.z.object({
-  /** Name of the skill that must have been called */
-  skillName: import_zod28.z.string().min(1)
-});
-var BuildPassedConfigSchema = import_zod28.z.object({
-  /** Command to run (default: "yarn build") */
-  command: import_zod28.z.string().optional(),
-  /** Expected exit code (default: 0) */
-  expectedExitCode: import_zod28.z.number().int().optional()
-});
-var LlmJudgeConfigSchema = import_zod28.z.object({
-  /** Prompt template; placeholders: {{output}}, {{cwd}}, {{changedFiles}}, {{trace}} */
-  prompt: import_zod28.z.string().min(1),
-  /** Optional system prompt for the judge */
-  systemPrompt: import_zod28.z.string().optional(),
-  /** Minimum score to pass (0-100, default 70) */
-  minScore: import_zod28.z.number().int().min(0).max(100).optional(),
-  /** Model for the judge (e.g. claude-3-5-haiku-20241022) */
-  model: import_zod28.z.string().optional(),
-  /** Max output tokens */
-  maxTokens: import_zod28.z.number().int().optional(),
-  /** Temperature (0-1) */
-  temperature: import_zod28.z.number().min(0).max(1).optional()
-});
-var AssertionConfigSchema = import_zod28.z.union([
-  SkillWasCalledConfigSchema,
-  BuildPassedConfigSchema,
-  LlmJudgeConfigSchema,
-  import_zod28.z.object({})
-  // Empty config for cases where defaults are used
-]);
-var CustomAssertionSchema = TenantEntitySchema.extend({
-  /** The assertion type */
-  type: AssertionTypeSchema,
-  /** Type-specific configuration */
-  config: AssertionConfigSchema
-});
-var CreateCustomAssertionInputSchema = CustomAssertionSchema.omit({
-  id: true,
-  createdAt: true,
-  updatedAt: true,
-  deleted: true
-});
-var UpdateCustomAssertionInputSchema = CreateCustomAssertionInputSchema.partial();
-function validateAssertionConfig(type, config) {
-  switch (type) {
-    case "skill_was_called":
-      return SkillWasCalledConfigSchema.safeParse(config).success;
-    case "build_passed":
-      return BuildPassedConfigSchema.safeParse(config).success;
-    case "llm_judge":
-    case "custom":
-      return LlmJudgeConfigSchema.safeParse(config).success;
-    default:
-      return false;
-  }
-}
-function getSkillWasCalledConfig(assertion) {
-  if (assertion.type !== "skill_was_called") return null;
-  const result = SkillWasCalledConfigSchema.safeParse(assertion.config);
-  return result.success ? result.data : null;
+// src/assertion/system-assertions.ts
+var SYSTEM_ASSERTION_IDS = {
+  SKILL_WAS_CALLED: "system:skill_was_called",
+  BUILD_PASSED: "system:build_passed"
+};
+function isSystemAssertionId(id) {
+  return id.startsWith("system:");
 }
-function getBuildPassedConfig(assertion) {
-  if (assertion.type !== "build_passed") return null;
-  const result = BuildPassedConfigSchema.safeParse(assertion.config);
-  return result.success ? result.data : null;
+var SYSTEM_ASSERTIONS = {
+  [SYSTEM_ASSERTION_IDS.SKILL_WAS_CALLED]: {
+    id: SYSTEM_ASSERTION_IDS.SKILL_WAS_CALLED,
+    name: "Skill Was Called",
+    description: "Check if a specific skill was invoked during the agent run",
+    type: "skill_was_called",
+    parameters: [
+      {
+        name: "skillName",
+        label: "Skill Name",
+        type: "string",
+        required: true
+      }
+    ]
+  },
+  [SYSTEM_ASSERTION_IDS.BUILD_PASSED]: {
+    id: SYSTEM_ASSERTION_IDS.BUILD_PASSED,
+    name: "Build Passed",
+    description: "Run a build command and verify it exits with expected code",
+    type: "build_passed",
+    parameters: [
+      {
+        name: "command",
+        label: "Build Command",
+        type: "string",
+        required: false,
+        defaultValue: "yarn build"
+      },
+      {
+        name: "expectedExitCode",
+        label: "Expected Exit Code",
+        type: "number",
+        required: false,
+        defaultValue: 0
+      },
+      {
+        name: "maxBuildTime",
+        label: "Max Build Time (ms)",
+        type: "number",
+        required: false
+      },
+      {
+        name: "maxMemory",
+        label: "Max Memory (MB)",
+        type: "number",
+        required: false
+      }
+    ]
+  }
+};
+function getSystemAssertions() {
+  return Object.values(SYSTEM_ASSERTIONS);
 }
-function getLlmJudgeConfig(assertion) {
-  if (assertion.type !== "llm_judge" && assertion.type !== "custom")
-    return null;
-  const result = LlmJudgeConfigSchema.safeParse(assertion.config);
-  return result.success ? result.data : null;
+function getSystemAssertion(id) {
+  return SYSTEM_ASSERTIONS[id];
 }
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
@@ -1147,6 +1261,8 @@ function getLlmJudgeConfig(assertion) {
   AllowedCommands,
   ApiCallSchema,
   AssertionConfigSchema,
+  AssertionParameterSchema,
+  AssertionParameterTypeSchema,
   AssertionResultSchema,
   AssertionResultStatus,
   AssertionSchema,
@@ -1213,6 +1329,9 @@ function getLlmJudgeConfig(assertion) {
   ProjectSchema,
   PromptResultSchema,
   SKILL_FOLDER_NAME_REGEX,
+  SYSTEM_ASSERTIONS,
+  SYSTEM_ASSERTION_IDS,
+  ScenarioAssertionLinkSchema,
   SiteConfigTestSchema,
   SkillMetadataSchema,
   SkillSchema,
@@ -1251,6 +1370,9 @@ function getLlmJudgeConfig(assertion) {
   getBuildPassedConfig,
   getLlmJudgeConfig,
   getSkillWasCalledConfig,
+  getSystemAssertion,
+  getSystemAssertions,
+  isSystemAssertionId,
   isValidSkillFolderName,
   parseTraceEventLine,
   validateAssertionConfig