@wix/evalforge-types 0.16.0 → 0.18.0

package/build/index.mjs

@@ -454,22 +454,145 @@ var EnvironmentSchema = z19.object({
 });
 
 // src/scenario/test-scenario.ts
+import { z as z21 } from "zod";
+
+// src/assertion/assertion.ts
 import { z as z20 } from "zod";
-var ExpectedFileSchema = z20.object({
+var AssertionTypeSchema = z20.enum([
+  "skill_was_called",
+  "build_passed",
+  "llm_judge"
+]);
+var AssertionParameterTypeSchema = z20.enum([
+  "string",
+  "number",
+  "boolean"
+]);
+var AssertionParameterSchema = z20.object({
+  /** Parameter name (used as key in params object) */
+  name: z20.string().min(1),
+  /** Display label for the parameter */
+  label: z20.string().min(1),
+  /** Parameter type */
+  type: AssertionParameterTypeSchema,
+  /** Whether this parameter is required */
+  required: z20.boolean(),
+  /** Default value (optional, used when not provided) */
+  defaultValue: z20.union([z20.string(), z20.number(), z20.boolean()]).optional()
+});
+var ScenarioAssertionLinkSchema = z20.object({
+  /** ID of the assertion (can be system assertion like 'system:skill_was_called' or custom assertion UUID) */
+  assertionId: z20.string(),
+  /** Parameter values for this assertion in this scenario */
+  params: z20.record(
+    z20.string(),
+    z20.union([z20.string(), z20.number(), z20.boolean(), z20.null()])
+  ).optional()
+});
+var SkillWasCalledConfigSchema = z20.object({
+  /** Name of the skill that must have been called */
+  skillName: z20.string().min(1)
+});
+var BuildPassedConfigSchema = z20.strictObject({
+  /** Command to run (default: "yarn build") */
+  command: z20.string().optional(),
+  /** Expected exit code (default: 0) */
+  expectedExitCode: z20.number().int().optional()
+});
+var LlmJudgeConfigSchema = z20.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: z20.string().min(1),
+  /** Optional system prompt for the judge */
+  systemPrompt: z20.string().optional(),
+  /** Minimum score to pass (0-100, default 70) */
+  minScore: z20.number().int().min(0).max(100).optional(),
+  /** Model for the judge (e.g. claude-3-5-haiku-20241022) */
+  model: z20.string().optional(),
+  /** Max output tokens */
+  maxTokens: z20.number().int().optional(),
+  /** Temperature (0-1) */
+  temperature: z20.number().min(0).max(1).optional(),
+  /** User-defined parameters for this assertion */
+  parameters: z20.array(AssertionParameterSchema).optional()
+});
+var AssertionConfigSchema = z20.union([
+  LlmJudgeConfigSchema,
+  // requires prompt - check first
+  SkillWasCalledConfigSchema,
+  // requires skillName
+  BuildPassedConfigSchema,
+  // all optional, uses strictObject to reject unknown keys
+  z20.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 = z21.object({
   /** Relative path where the file should be created */
-  path: z20.string(),
+  path: z21.string(),
   /** Optional expected content */
-  content: z20.string().optional()
+  content: z21.string().optional()
 });
 var TestScenarioSchema = TenantEntitySchema.extend({
   /** The prompt sent to the agent to trigger the task */
-  triggerPrompt: z20.string().min(10),
+  triggerPrompt: z21.string().min(10),
   /** ID of the template to use for this scenario (null = no template) */
-  templateId: z20.string().nullish(),
+  templateId: z21.string().nullish(),
   /** Inline assertions to evaluate for this scenario (legacy) */
-  assertions: z20.array(AssertionSchema).optional(),
-  /** IDs of saved assertions to evaluate (from assertions table) */
-  assertionIds: z20.array(z20.string()).optional()
+  assertions: z21.array(AssertionSchema).optional(),
+  /** IDs of saved assertions to evaluate (from assertions table) - legacy, use assertionLinks */
+  assertionIds: z21.array(z21.string()).optional(),
+  /** Linked assertions with per-scenario parameter values */
+  assertionLinks: z21.array(ScenarioAssertionLinkSchema).optional()
 });
 var CreateTestScenarioInputSchema = TestScenarioSchema.omit({
   id: true,
@@ -480,10 +603,10 @@ var CreateTestScenarioInputSchema = TestScenarioSchema.omit({
 var UpdateTestScenarioInputSchema = CreateTestScenarioInputSchema.partial();
 
 // src/suite/test-suite.ts
-import { z as z21 } from "zod";
+import { z as z22 } from "zod";
 var TestSuiteSchema = TenantEntitySchema.extend({
   /** IDs of test scenarios in this suite */
-  scenarioIds: z21.array(z21.string())
+  scenarioIds: z22.array(z22.string())
 });
 var CreateTestSuiteInputSchema = TestSuiteSchema.omit({
   id: true,
@@ -494,21 +617,21 @@ var CreateTestSuiteInputSchema = TestSuiteSchema.omit({
 var UpdateTestSuiteInputSchema = CreateTestSuiteInputSchema.partial();
 
 // src/evaluation/metrics.ts
-import { z as z22 } from "zod";
-var TokenUsageSchema = z22.object({
-  prompt: z22.number(),
-  completion: z22.number(),
-  total: z22.number()
-});
-var EvalMetricsSchema = z22.object({
-  totalAssertions: z22.number(),
-  passed: z22.number(),
-  failed: z22.number(),
-  skipped: z22.number(),
-  errors: z22.number(),
-  passRate: z22.number(),
-  avgDuration: z22.number(),
-  totalDuration: z22.number()
+import { z as z23 } from "zod";
+var TokenUsageSchema = z23.object({
+  prompt: z23.number(),
+  completion: z23.number(),
+  total: z23.number()
+});
+var EvalMetricsSchema = z23.object({
+  totalAssertions: z23.number(),
+  passed: z23.number(),
+  failed: z23.number(),
+  skipped: z23.number(),
+  errors: z23.number(),
+  passRate: z23.number(),
+  avgDuration: z23.number(),
+  totalDuration: z23.number()
 });
 var EvalStatus = /* @__PURE__ */ ((EvalStatus2) => {
   EvalStatus2["PENDING"] = "pending";
@@ -518,7 +641,7 @@ var EvalStatus = /* @__PURE__ */ ((EvalStatus2) => {
   EvalStatus2["CANCELLED"] = "cancelled";
   return EvalStatus2;
 })(EvalStatus || {});
-var EvalStatusSchema = z22.enum(EvalStatus);
+var EvalStatusSchema = z23.enum(EvalStatus);
 var LLMStepType = /* @__PURE__ */ ((LLMStepType2) => {
   LLMStepType2["COMPLETION"] = "completion";
   LLMStepType2["TOOL_USE"] = "tool_use";
@@ -526,52 +649,52 @@ var LLMStepType = /* @__PURE__ */ ((LLMStepType2) => {
   LLMStepType2["THINKING"] = "thinking";
   return LLMStepType2;
 })(LLMStepType || {});
-var LLMTraceStepSchema = z22.object({
-  id: z22.string(),
-  stepNumber: z22.number(),
-  type: z22.enum(LLMStepType),
-  model: z22.string(),
-  provider: z22.string(),
-  startedAt: z22.string(),
-  durationMs: z22.number(),
+var LLMTraceStepSchema = z23.object({
+  id: z23.string(),
+  stepNumber: z23.number(),
+  type: z23.enum(LLMStepType),
+  model: z23.string(),
+  provider: z23.string(),
+  startedAt: z23.string(),
+  durationMs: z23.number(),
   tokenUsage: TokenUsageSchema,
-  costUsd: z22.number(),
-  toolName: z22.string().optional(),
-  toolArguments: z22.string().optional(),
-  inputPreview: z22.string().optional(),
-  outputPreview: z22.string().optional(),
-  success: z22.boolean(),
-  error: z22.string().optional()
-});
-var LLMBreakdownStatsSchema = z22.object({
-  count: z22.number(),
-  durationMs: z22.number(),
-  tokens: z22.number(),
-  costUsd: z22.number()
-});
-var LLMTraceSummarySchema = z22.object({
-  totalSteps: z22.number(),
-  totalDurationMs: z22.number(),
+  costUsd: z23.number(),
+  toolName: z23.string().optional(),
+  toolArguments: z23.string().optional(),
+  inputPreview: z23.string().optional(),
+  outputPreview: z23.string().optional(),
+  success: z23.boolean(),
+  error: z23.string().optional()
+});
+var LLMBreakdownStatsSchema = z23.object({
+  count: z23.number(),
+  durationMs: z23.number(),
+  tokens: z23.number(),
+  costUsd: z23.number()
+});
+var LLMTraceSummarySchema = z23.object({
+  totalSteps: z23.number(),
+  totalDurationMs: z23.number(),
   totalTokens: TokenUsageSchema,
-  totalCostUsd: z22.number(),
-  stepTypeBreakdown: z22.record(z22.string(), LLMBreakdownStatsSchema).optional(),
-  modelBreakdown: z22.record(z22.string(), LLMBreakdownStatsSchema),
-  modelsUsed: z22.array(z22.string())
-});
-var LLMTraceSchema = z22.object({
-  id: z22.string(),
-  steps: z22.array(LLMTraceStepSchema),
+  totalCostUsd: z23.number(),
+  stepTypeBreakdown: z23.record(z23.string(), LLMBreakdownStatsSchema).optional(),
+  modelBreakdown: z23.record(z23.string(), LLMBreakdownStatsSchema),
+  modelsUsed: z23.array(z23.string())
+});
+var LLMTraceSchema = z23.object({
+  id: z23.string(),
+  steps: z23.array(LLMTraceStepSchema),
   summary: LLMTraceSummarySchema
 });
 
 // src/evaluation/eval-result.ts
-import { z as z25 } from "zod";
+import { z as z26 } from "zod";
 
 // src/evaluation/eval-run.ts
-import { z as z24 } from "zod";
+import { z as z25 } from "zod";
 
 // src/evaluation/live-trace.ts
-import { z as z23 } from "zod";
+import { z as z24 } from "zod";
 var LiveTraceEventType = /* @__PURE__ */ ((LiveTraceEventType2) => {
   LiveTraceEventType2["THINKING"] = "thinking";
   LiveTraceEventType2["TOOL_USE"] = "tool_use";
@@ -585,37 +708,37 @@ var LiveTraceEventType = /* @__PURE__ */ ((LiveTraceEventType2) => {
   LiveTraceEventType2["USER"] = "user";
   return LiveTraceEventType2;
 })(LiveTraceEventType || {});
-var LiveTraceEventSchema = z23.object({
+var LiveTraceEventSchema = z24.object({
   /** The evaluation run ID */
-  evalRunId: z23.string(),
+  evalRunId: z24.string(),
   /** The scenario ID being executed */
-  scenarioId: z23.string(),
+  scenarioId: z24.string(),
   /** The scenario name for display */
-  scenarioName: z23.string(),
+  scenarioName: z24.string(),
   /** The target ID (skill, agent, etc.) */
-  targetId: z23.string(),
+  targetId: z24.string(),
   /** The target name for display */
-  targetName: z23.string(),
+  targetName: z24.string(),
   /** Step number in the current scenario execution */
-  stepNumber: z23.number(),
+  stepNumber: z24.number(),
   /** Type of trace event */
-  type: z23.enum(LiveTraceEventType),
+  type: z24.enum(LiveTraceEventType),
   /** Tool name if this is a tool_use event */
-  toolName: z23.string().optional(),
+  toolName: z24.string().optional(),
   /** Tool arguments preview (truncated JSON) */
-  toolArgs: z23.string().optional(),
+  toolArgs: z24.string().optional(),
   /** Output preview (truncated text) */
-  outputPreview: z23.string().optional(),
+  outputPreview: z24.string().optional(),
   /** File path for file operations */
-  filePath: z23.string().optional(),
+  filePath: z24.string().optional(),
   /** Elapsed time in milliseconds for progress events */
-  elapsedMs: z23.number().optional(),
+  elapsedMs: z24.number().optional(),
   /** Thinking/reasoning text from Claude */
-  thinking: z23.string().optional(),
+  thinking: z24.string().optional(),
   /** Timestamp when this event occurred */
-  timestamp: z23.string(),
+  timestamp: z24.string(),
   /** Whether this is the final event for this scenario */
-  isComplete: z23.boolean()
+  isComplete: z24.boolean()
 });
 var TRACE_EVENT_PREFIX = "TRACE_EVENT:";
 function parseTraceEventLine(line) {
@@ -643,14 +766,14 @@ var TriggerType = /* @__PURE__ */ ((TriggerType2) => {
   TriggerType2["MANUAL"] = "MANUAL";
   return TriggerType2;
 })(TriggerType || {});
-var TriggerMetadataSchema = z24.object({
-  version: z24.string().optional(),
-  resourceUpdated: z24.array(z24.string()).optional()
+var TriggerMetadataSchema = z25.object({
+  version: z25.string().optional(),
+  resourceUpdated: z25.array(z25.string()).optional()
 });
-var TriggerSchema = z24.object({
-  id: z24.string(),
+var TriggerSchema = z25.object({
+  id: z25.string(),
   metadata: TriggerMetadataSchema.optional(),
-  type: z24.enum(TriggerType)
+  type: z25.enum(TriggerType)
 });
 var FailureCategory = /* @__PURE__ */ ((FailureCategory2) => {
   FailureCategory2["MISSING_FILE"] = "missing_file";
@@ -668,28 +791,28 @@ var FailureSeverity = /* @__PURE__ */ ((FailureSeverity2) => {
   FailureSeverity2["LOW"] = "low";
   return FailureSeverity2;
 })(FailureSeverity || {});
-var DiffLineTypeSchema = z24.enum(["added", "removed", "unchanged"]);
-var DiffLineSchema = z24.object({
+var DiffLineTypeSchema = z25.enum(["added", "removed", "unchanged"]);
+var DiffLineSchema = z25.object({
   type: DiffLineTypeSchema,
-  content: z24.string(),
-  lineNumber: z24.number()
-});
-var DiffContentSchema = z24.object({
-  path: z24.string(),
-  expected: z24.string(),
-  actual: z24.string(),
-  diffLines: z24.array(DiffLineSchema),
-  renamedFrom: z24.string().optional()
-});
-var CommandExecutionSchema = z24.object({
-  command: z24.string(),
-  exitCode: z24.number(),
-  output: z24.string().optional(),
-  duration: z24.number()
-});
-var FileModificationSchema = z24.object({
-  path: z24.string(),
-  action: z24.enum(["created", "modified", "deleted"])
+  content: z25.string(),
+  lineNumber: z25.number()
+});
+var DiffContentSchema = z25.object({
+  path: z25.string(),
+  expected: z25.string(),
+  actual: z25.string(),
+  diffLines: z25.array(DiffLineSchema),
+  renamedFrom: z25.string().optional()
+});
+var CommandExecutionSchema = z25.object({
+  command: z25.string(),
+  exitCode: z25.number(),
+  output: z25.string().optional(),
+  duration: z25.number()
+});
+var FileModificationSchema = z25.object({
+  path: z25.string(),
+  action: z25.enum(["created", "modified", "deleted"])
 });
 var TemplateFileStatus = /* @__PURE__ */ ((TemplateFileStatus2) => {
   TemplateFileStatus2["NEW"] = "new";
@@ -697,75 +820,75 @@ var TemplateFileStatus = /* @__PURE__ */ ((TemplateFileStatus2) => {
   TemplateFileStatus2["UNCHANGED"] = "unchanged";
   return TemplateFileStatus2;
 })(TemplateFileStatus || {});
-var TemplateFileSchema = z24.object({
+var TemplateFileSchema = z25.object({
   /** Relative path within the template */
-  path: z24.string(),
+  path: z25.string(),
   /** Full file content after execution */
-  content: z24.string(),
+  content: z25.string(),
   /** File status (new, modified, unchanged) */
-  status: z24.enum(["new", "modified", "unchanged"])
-});
-var ApiCallSchema = z24.object({
-  endpoint: z24.string(),
-  tokensUsed: z24.number(),
-  duration: z24.number()
-});
-var ExecutionTraceSchema = z24.object({
-  commands: z24.array(CommandExecutionSchema),
-  filesModified: z24.array(FileModificationSchema),
-  apiCalls: z24.array(ApiCallSchema),
-  totalDuration: z24.number()
-});
-var FailureAnalysisSchema = z24.object({
-  category: z24.enum(FailureCategory),
-  severity: z24.enum(FailureSeverity),
-  summary: z24.string(),
-  details: z24.string(),
-  rootCause: z24.string(),
-  suggestedFix: z24.string(),
-  relatedAssertions: z24.array(z24.string()),
-  codeSnippet: z24.string().optional(),
-  similarIssues: z24.array(z24.string()).optional(),
-  patternId: z24.string().optional(),
+  status: z25.enum(["new", "modified", "unchanged"])
+});
+var ApiCallSchema = z25.object({
+  endpoint: z25.string(),
+  tokensUsed: z25.number(),
+  duration: z25.number()
+});
+var ExecutionTraceSchema = z25.object({
+  commands: z25.array(CommandExecutionSchema),
+  filesModified: z25.array(FileModificationSchema),
+  apiCalls: z25.array(ApiCallSchema),
+  totalDuration: z25.number()
+});
+var FailureAnalysisSchema = z25.object({
+  category: z25.enum(FailureCategory),
+  severity: z25.enum(FailureSeverity),
+  summary: z25.string(),
+  details: z25.string(),
+  rootCause: z25.string(),
+  suggestedFix: z25.string(),
+  relatedAssertions: z25.array(z25.string()),
+  codeSnippet: z25.string().optional(),
+  similarIssues: z25.array(z25.string()).optional(),
+  patternId: z25.string().optional(),
   // Extended fields for detailed debugging
   diff: DiffContentSchema.optional(),
   executionTrace: ExecutionTraceSchema.optional()
 });
 var EvalRunSchema = TenantEntitySchema.extend({
   /** Agent ID for this run */
-  agentId: z24.string().optional(),
+  agentId: z25.string().optional(),
   /** Skills group ID for this run */
-  skillsGroupId: z24.string().optional(),
+  skillsGroupId: z25.string().optional(),
   /** Scenario IDs to run */
-  scenarioIds: z24.array(z24.string()),
+  scenarioIds: z25.array(z25.string()),
   /** Current status */
   status: EvalStatusSchema,
   /** Progress percentage (0-100) */
-  progress: z24.number(),
+  progress: z25.number(),
   /** Results for each scenario/target combination (lazy to break eval-result ↔ eval-run cycle) */
-  results: z24.array(z24.lazy(() => EvalRunResultSchema)),
+  results: z25.array(z25.lazy(() => EvalRunResultSchema)),
   /** Aggregated metrics across all results */
   aggregateMetrics: EvalMetricsSchema,
   /** Failure analyses */
-  failureAnalyses: z24.array(FailureAnalysisSchema).optional(),
+  failureAnalyses: z25.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: z24.string().optional(),
+  startedAt: z25.string().optional(),
   /** When the run completed */
-  completedAt: z24.string().optional(),
+  completedAt: z25.string().optional(),
   /** Live trace events captured during execution (for playback on results page) */
-  liveTraceEvents: z24.array(LiveTraceEventSchema).optional(),
+  liveTraceEvents: z25.array(LiveTraceEventSchema).optional(),
   /** Remote job ID for tracking execution in Dev Machines */
-  jobId: z24.string().optional(),
+  jobId: z25.string().optional(),
   /** Remote job status from the Dev Machine API (PENDING, RUNNING, COMPLETED, FAILED, CANCELLED) */
-  jobStatus: z24.string().optional(),
+  jobStatus: z25.string().optional(),
   /** Remote job error message if the job failed */
-  jobError: z24.string().optional(),
+  jobError: z25.string().optional(),
   /** Timestamp of the last job status check */
-  jobStatusCheckedAt: z24.string().optional()
+  jobStatusCheckedAt: z25.string().optional()
 });
 var CreateEvalRunInputSchema = EvalRunSchema.omit({
   id: true,
@@ -778,28 +901,28 @@ var CreateEvalRunInputSchema = EvalRunSchema.omit({
   startedAt: true,
   completedAt: true
 });
-var EvaluationProgressSchema = z24.object({
-  runId: z24.string(),
-  targetId: z24.string(),
-  totalScenarios: z24.number(),
-  completedScenarios: z24.number(),
-  scenarioProgress: z24.array(
-    z24.object({
-      scenarioId: z24.string(),
-      currentStep: z24.string(),
-      error: z24.string().optional()
+var EvaluationProgressSchema = z25.object({
+  runId: z25.string(),
+  targetId: z25.string(),
+  totalScenarios: z25.number(),
+  completedScenarios: z25.number(),
+  scenarioProgress: z25.array(
+    z25.object({
+      scenarioId: z25.string(),
+      currentStep: z25.string(),
+      error: z25.string().optional()
     })
   ),
-  createdAt: z24.number()
+  createdAt: z25.number()
 });
-var EvaluationLogSchema = z24.object({
-  runId: z24.string(),
-  scenarioId: z24.string(),
-  log: z24.object({
-    level: z24.enum(["info", "error", "debug"]),
-    message: z24.string().optional(),
-    args: z24.array(z24.any()).optional(),
-    error: z24.string().optional()
+var EvaluationLogSchema = z25.object({
+  runId: z25.string(),
+  scenarioId: z25.string(),
+  log: z25.object({
+    level: z25.enum(["info", "error", "debug"]),
+    message: z25.string().optional(),
+    args: z25.array(z25.any()).optional(),
+    error: z25.string().optional()
   })
 });
 var LLM_TIMEOUT = 12e4;
@@ -812,91 +935,91 @@ var AssertionResultStatus = /* @__PURE__ */ ((AssertionResultStatus2) => {
   AssertionResultStatus2["ERROR"] = "error";
   return AssertionResultStatus2;
 })(AssertionResultStatus || {});
-var AssertionResultSchema = z25.object({
-  id: z25.string(),
-  assertionId: z25.string(),
-  assertionType: z25.string(),
-  assertionName: z25.string(),
-  status: z25.enum(AssertionResultStatus),
-  message: z25.string().optional(),
-  expected: z25.string().optional(),
-  actual: z25.string().optional(),
-  duration: z25.number().optional(),
-  details: z25.record(z25.string(), z25.unknown()).optional(),
-  llmTraceSteps: z25.array(LLMTraceStepSchema).optional()
-});
-var EvalRunResultSchema = z25.object({
-  id: z25.string(),
-  targetId: z25.string(),
-  targetName: z25.string().optional(),
-  scenarioId: z25.string(),
-  scenarioName: z25.string(),
+var AssertionResultSchema = z26.object({
+  id: z26.string(),
+  assertionId: z26.string(),
+  assertionType: z26.string(),
+  assertionName: z26.string(),
+  status: z26.enum(AssertionResultStatus),
+  message: z26.string().optional(),
+  expected: z26.string().optional(),
+  actual: z26.string().optional(),
+  duration: z26.number().optional(),
+  details: z26.record(z26.string(), z26.unknown()).optional(),
+  llmTraceSteps: z26.array(LLMTraceStepSchema).optional()
+});
+var EvalRunResultSchema = z26.object({
+  id: z26.string(),
+  targetId: z26.string(),
+  targetName: z26.string().optional(),
+  scenarioId: z26.string(),
+  scenarioName: z26.string(),
   modelConfig: ModelConfigSchema.optional(),
-  assertionResults: z25.array(AssertionResultSchema),
+  assertionResults: z26.array(AssertionResultSchema),
   metrics: EvalMetricsSchema.optional(),
-  passed: z25.number(),
-  failed: z25.number(),
-  passRate: z25.number(),
-  duration: z25.number(),
-  outputText: z25.string().optional(),
-  files: z25.array(ExpectedFileSchema).optional(),
-  fileDiffs: z25.array(DiffContentSchema).optional(),
+  passed: z26.number(),
+  failed: z26.number(),
+  passRate: z26.number(),
+  duration: z26.number(),
+  outputText: z26.string().optional(),
+  files: z26.array(ExpectedFileSchema).optional(),
+  fileDiffs: z26.array(DiffContentSchema).optional(),
   /** Full template files after execution with status indicators */
-  templateFiles: z25.array(TemplateFileSchema).optional(),
-  startedAt: z25.string().optional(),
-  completedAt: z25.string().optional(),
+  templateFiles: z26.array(TemplateFileSchema).optional(),
+  startedAt: z26.string().optional(),
+  completedAt: z26.string().optional(),
   llmTrace: LLMTraceSchema.optional()
 });
-var PromptResultSchema = z25.object({
-  text: z25.string(),
-  files: z25.array(z25.unknown()).optional(),
-  finishReason: z25.string().optional(),
-  reasoning: z25.string().optional(),
-  reasoningDetails: z25.unknown().optional(),
-  toolCalls: z25.array(z25.unknown()).optional(),
-  toolResults: z25.array(z25.unknown()).optional(),
-  warnings: z25.array(z25.unknown()).optional(),
-  sources: z25.array(z25.unknown()).optional(),
-  steps: z25.array(z25.unknown()),
-  generationTimeMs: z25.number(),
-  prompt: z25.string(),
-  systemPrompt: z25.string(),
-  usage: z25.object({
-    totalTokens: z25.number().optional(),
-    totalMicrocentsSpent: z25.number().optional()
+var PromptResultSchema = z26.object({
+  text: z26.string(),
+  files: z26.array(z26.unknown()).optional(),
+  finishReason: z26.string().optional(),
+  reasoning: z26.string().optional(),
+  reasoningDetails: z26.unknown().optional(),
+  toolCalls: z26.array(z26.unknown()).optional(),
+  toolResults: z26.array(z26.unknown()).optional(),
+  warnings: z26.array(z26.unknown()).optional(),
+  sources: z26.array(z26.unknown()).optional(),
+  steps: z26.array(z26.unknown()),
+  generationTimeMs: z26.number(),
+  prompt: z26.string(),
+  systemPrompt: z26.string(),
+  usage: z26.object({
+    totalTokens: z26.number().optional(),
+    totalMicrocentsSpent: z26.number().optional()
   })
 });
-var EvaluationResultSchema = z25.object({
-  id: z25.string(),
-  runId: z25.string(),
-  timestamp: z25.number(),
+var EvaluationResultSchema = z26.object({
+  id: z26.string(),
+  runId: z26.string(),
+  timestamp: z26.number(),
   promptResult: PromptResultSchema,
-  testResults: z25.array(z25.unknown()),
-  tags: z25.array(z25.string()).optional(),
-  feedback: z25.string().optional(),
-  score: z25.number(),
-  suiteId: z25.string().optional()
-});
-var LeanEvaluationResultSchema = z25.object({
-  id: z25.string(),
-  runId: z25.string(),
-  timestamp: z25.number(),
-  tags: z25.array(z25.string()).optional(),
-  scenarioId: z25.string(),
-  scenarioVersion: z25.number().optional(),
-  targetId: z25.string(),
-  targetVersion: z25.number().optional(),
-  suiteId: z25.string().optional(),
-  score: z25.number(),
-  time: z25.number().optional(),
-  microcentsSpent: z25.number().optional()
+  testResults: z26.array(z26.unknown()),
+  tags: z26.array(z26.string()).optional(),
+  feedback: z26.string().optional(),
+  score: z26.number(),
+  suiteId: z26.string().optional()
+});
+var LeanEvaluationResultSchema = z26.object({
+  id: z26.string(),
+  runId: z26.string(),
+  timestamp: z26.number(),
+  tags: z26.array(z26.string()).optional(),
+  scenarioId: z26.string(),
+  scenarioVersion: z26.number().optional(),
+  targetId: z26.string(),
+  targetVersion: z26.number().optional(),
+  suiteId: z26.string().optional(),
+  score: z26.number(),
+  time: z26.number().optional(),
+  microcentsSpent: z26.number().optional()
 });
 
 // src/project/project.ts
-import { z as z26 } from "zod";
+import { z as z27 } from "zod";
 var ProjectSchema = BaseEntitySchema.extend({
-  appId: z26.string().optional().describe("The ID of the app in Dev Center"),
-  appSecret: z26.string().optional().describe("The secret of the app in Dev Center")
+  appId: z27.string().optional().describe("The ID of the app in Dev Center"),
+  appSecret: z27.string().optional().describe("The secret of the app in Dev Center")
 });
 var CreateProjectInputSchema = ProjectSchema.omit({
   id: true,
@@ -907,10 +1030,10 @@ var CreateProjectInputSchema = ProjectSchema.omit({
 var UpdateProjectInputSchema = CreateProjectInputSchema.partial();
 
 // src/template/template.ts
-import { z as z27 } from "zod";
+import { z as z28 } from "zod";
 var TemplateSchema = TenantEntitySchema.extend({
   /** URL to download the template from */
-  downloadUrl: z27.url()
+  downloadUrl: z28.url()
 });
 var CreateTemplateInputSchema = TemplateSchema.omit({
   id: true,
@@ -920,86 +1043,69 @@ var CreateTemplateInputSchema = TemplateSchema.omit({
 });
 var UpdateTemplateInputSchema = CreateTemplateInputSchema.partial();
 
-// src/assertion/assertion.ts
-import { z as z28 } from "zod";
-var AssertionTypeSchema = z28.enum([
-  "skill_was_called",
-  "build_passed",
-  "llm_judge",
-  "custom"
-]);
-var SkillWasCalledConfigSchema = z28.object({
-  /** Name of the skill that must have been called */
-  skillName: z28.string().min(1)
-});
-var BuildPassedConfigSchema = z28.object({
-  /** Command to run (default: "yarn build") */
-  command: z28.string().optional(),
-  /** Expected exit code (default: 0) */
-  expectedExitCode: z28.number().int().optional()
-});
-var LlmJudgeConfigSchema = z28.object({
-  /** Prompt template; placeholders: {{output}}, {{cwd}}, {{changedFiles}}, {{trace}} */
-  prompt: z28.string().min(1),
-  /** Optional system prompt for the judge */
-  systemPrompt: z28.string().optional(),
-  /** Minimum score to pass (0-100, default 70) */
-  minScore: z28.number().int().min(0).max(100).optional(),
-  /** Model for the judge (e.g. claude-3-5-haiku-20241022) */
-  model: z28.string().optional(),
-  /** Max output tokens */
-  maxTokens: z28.number().int().optional(),
-  /** Temperature (0-1) */
-  temperature: z28.number().min(0).max(1).optional()
-});
-var AssertionConfigSchema = z28.union([
-  SkillWasCalledConfigSchema,
-  BuildPassedConfigSchema,
-  LlmJudgeConfigSchema,
-  z28.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];
 }
 export {
   AVAILABLE_MODELS,
@@ -1008,6 +1114,8 @@ export {
   AllowedCommands,
   ApiCallSchema,
   AssertionConfigSchema,
+  AssertionParameterSchema,
+  AssertionParameterTypeSchema,
   AssertionResultSchema,
   AssertionResultStatus,
   AssertionSchema,
@@ -1074,6 +1182,9 @@ export {
   ProjectSchema,
   PromptResultSchema,
   SKILL_FOLDER_NAME_REGEX,
+  SYSTEM_ASSERTIONS,
+  SYSTEM_ASSERTION_IDS,
+  ScenarioAssertionLinkSchema,
   SiteConfigTestSchema,
   SkillMetadataSchema,
   SkillSchema,
@@ -1112,6 +1223,9 @@ export {
   getBuildPassedConfig,
   getLlmJudgeConfig,
   getSkillWasCalledConfig,
+  getSystemAssertion,
+  getSystemAssertions,
+  isSystemAssertionId,
   isValidSkillFolderName,
   parseTraceEventLine,
   validateAssertionConfig