@wix/evalforge-types 0.41.0 → 0.42.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/build/index.js CHANGED
@@ -61,6 +61,7 @@ __export(index_exports, {
61
61
  CreateTestScenarioInputSchema: () => CreateTestScenarioInputSchema,
62
62
  CreateTestSuiteInputSchema: () => CreateTestSuiteInputSchema,
63
63
  CustomAssertionSchema: () => CustomAssertionSchema,
64
+ DEFAULT_EVALUATOR_SYSTEM_PROMPT: () => DEFAULT_EVALUATOR_SYSTEM_PROMPT,
64
65
  DEFAULT_JUDGE_MODEL: () => DEFAULT_JUDGE_MODEL,
65
66
  DiffContentSchema: () => DiffContentSchema,
66
67
  DiffLineSchema: () => DiffLineSchema,
@@ -1170,6 +1171,15 @@ var TargetSchema = TenantEntitySchema.extend({
1170
1171
 
1171
1172
  // src/target/agent.ts
1172
1173
  var import_zod6 = require("zod");
1174
+ var DEFAULT_EVALUATOR_SYSTEM_PROMPT = `IMPORTANT: This is an automated evaluation run. Follow these guidelines:
1175
+ 1. Execute the requested changes immediately without asking for confirmation.
1176
+ 2. Do NOT ask "would you like me to proceed?" or similar questions.
1177
+ 3. Do NOT use the Task tool to delegate simple operations - do them directly yourself.
1178
+ 4. Keep your approach simple and direct - avoid excessive planning.
1179
+ 5. Make targeted edits using Read and Edit tools rather than exploring the entire codebase.
1180
+ 6. If you encounter an error, fix it directly rather than starting over.
1181
+ 7. Your project root is the current working directory. Always create and modify source code files relative to the project root, NOT inside .claude/skills/ directories.
1182
+ 8. Before finishing, run the project's package manager install command (e.g. \`npm install\`, \`yarn install\`, or \`pnpm install\` depending on the lockfile present) to ensure all dependencies are installed and the project is ready to build.`;
1173
1183
  var AgentRunCommand = /* @__PURE__ */ ((AgentRunCommand2) => {
1174
1184
  AgentRunCommand2["CLAUDE"] = "claude";
1175
1185
  return AgentRunCommand2;
@@ -1183,7 +1193,10 @@ var AgentSchema = TargetSchema.extend({
1183
1193
  /** Command to run the agent */
1184
1194
  runCommand: AgentRunCommandSchema,
1185
1195
  /** Optional model configuration override */
1186
- modelConfig: ModelConfigSchema.optional()
1196
+ modelConfig: ModelConfigSchema.optional(),
1197
+ systemPrompt: import_zod6.z.string().nullish().describe(
1198
+ "Override for eval runs. undefined=default instructions, null=raw agent, string=append to claude_code preset. See https://docs.anthropic.com/en/docs/claude-code/sdk/modifying-system-prompts"
1199
+ )
1187
1200
  });
1188
1201
  var CreateAgentInputSchema = AgentSchema.omit({
1189
1202
  id: true,
@@ -1192,7 +1205,8 @@ var CreateAgentInputSchema = AgentSchema.omit({
1192
1205
  deleted: true
1193
1206
  });
1194
1207
  var UpdateAgentInputSchema = CreateAgentInputSchema.partial().extend({
1195
- modelConfig: ModelConfigSchema.optional().nullable()
1208
+ modelConfig: ModelConfigSchema.optional().nullable(),
1209
+ systemPrompt: import_zod6.z.string().optional().nullable()
1196
1210
  });
1197
1211
 
1198
1212
  // src/target/skill.ts
@@ -2399,6 +2413,7 @@ function getSystemAssertion(id) {
2399
2413
  CreateTestScenarioInputSchema,
2400
2414
  CreateTestSuiteInputSchema,
2401
2415
  CustomAssertionSchema,
2416
+ DEFAULT_EVALUATOR_SYSTEM_PROMPT,
2402
2417
  DEFAULT_JUDGE_MODEL,
2403
2418
  DiffContentSchema,
2404
2419
  DiffLineSchema,