Skip to content

fix(evals): update eval tests for invoke_agent telemetry and project-scoped memory #25502

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
SandyTao520 merged 1 commit into from
Apr 15, 2026
Merged

fix(evals): update eval tests for invoke_agent telemetry and project-scoped memory #25502

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 11 additions & 3 deletions evals/cli_help_delegation.eval.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,9 +17,17 @@ describe('CliHelpAgent Delegation', () => {
timeout: 60000,
assert: async (rig, _result) => {
const toolLogs = rig.readToolLogs();
const toolCallIndex = toolLogs.findIndex(
(log) => log.toolRequest.name === 'cli_help',
);
const toolCallIndex = toolLogs.findIndex((log) => {
if (log.toolRequest.name === 'invoke_agent') {
try {
const args = JSON.parse(log.toolRequest.args);
return args.agent_name === 'cli_help';
} catch {
return false;
}
}
return false;
});
expect(toolCallIndex).toBeGreaterThan(-1);
expect(toolCallIndex).toBeLessThan(5); // Called within first 5 turns
},
Expand Down
17 changes: 14 additions & 3 deletions evals/generalist_agent.eval.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,11 +26,22 @@ describe('generalist_agent', () => {
prompt:
'Please use the generalist agent to create a file called "generalist_test_file.txt" containing exactly the following text: success',
assert: async (rig) => {
// 1) Verify the generalist agent was invoked
const foundToolCall = await rig.waitForToolCall('generalist');
// 1) Verify the generalist agent was invoked via invoke_agent
const foundToolCall = await rig.waitForToolCall(
'invoke_agent',
undefined,
(args) => {
try {
const parsed = JSON.parse(args);
return parsed.agent_name === 'generalist';
} catch {
return false;
}
},
);
expect(
foundToolCall,
'Expected to find a tool call for generalist agent',
'Expected to find an invoke_agent tool call for generalist agent',
).toBeTruthy();

// 2) Verify the file was created as expected
Expand Down
94 changes: 63 additions & 31 deletions evals/save_memory.eval.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,22 +145,30 @@ describe('save_memory', () => {
},
});

const ignoringDbSchemaLocation =
"Agent ignores workspace's database schema location";
const savingDbSchemaLocationAsProjectMemory =
'Agent saves workspace database schema location as project memory';
evalTest('USUALLY_PASSES', {
suiteName: 'default',
suiteType: 'behavioral',
name: ignoringDbSchemaLocation,
name: savingDbSchemaLocationAsProjectMemory,
prompt: `The database schema for this workspace is located in \`db/schema.sql\`.`,
assert: async (rig, result) => {
await rig.waitForTelemetryReady();
const wasToolCalled = rig
.readToolLogs()
.some((log) => log.toolRequest.name === 'save_memory');
const wasToolCalled = await rig.waitForToolCall(
'save_memory',
undefined,
(args) => {
try {
const params = JSON.parse(args);
return params.scope === 'project';
} catch {
return false;
}
},
);
expect(
wasToolCalled,
'save_memory should not be called for workspace-specific information',
).toBe(false);
'Expected save_memory to be called with scope="project" for workspace-specific information',
).toBe(true);

assertModelHasOutput(result);
},
Expand Down Expand Up @@ -188,42 +196,59 @@ describe('save_memory', () => {
},
});

const ignoringBuildArtifactLocation =
'Agent ignores workspace build artifact location';
const savingBuildArtifactLocationAsProjectMemory =
'Agent saves workspace build artifact location as project memory';
evalTest('USUALLY_PASSES', {
suiteName: 'default',
suiteType: 'behavioral',
name: ignoringBuildArtifactLocation,
name: savingBuildArtifactLocationAsProjectMemory,
prompt: `In this workspace, build artifacts are stored in the \`dist/artifacts\` directory.`,
assert: async (rig, result) => {
await rig.waitForTelemetryReady();
const wasToolCalled = rig
.readToolLogs()
.some((log) => log.toolRequest.name === 'save_memory');
const wasToolCalled = await rig.waitForToolCall(
'save_memory',
undefined,
(args) => {
try {
const params = JSON.parse(args);
return params.scope === 'project';
} catch {
return false;
}
},
);
expect(
wasToolCalled,
'save_memory should not be called for workspace-specific information',
).toBe(false);
'Expected save_memory to be called with scope="project" for workspace-specific information',
).toBe(true);

assertModelHasOutput(result);
},
});

const ignoringMainEntryPoint = "Agent ignores workspace's main entry point";
const savingMainEntryPointAsProjectMemory =
'Agent saves workspace main entry point as project memory';
evalTest('USUALLY_PASSES', {
suiteName: 'default',
suiteType: 'behavioral',
name: ignoringMainEntryPoint,
name: savingMainEntryPointAsProjectMemory,
prompt: `The main entry point for this workspace is \`src/index.js\`.`,
assert: async (rig, result) => {
await rig.waitForTelemetryReady();
const wasToolCalled = rig
.readToolLogs()
.some((log) => log.toolRequest.name === 'save_memory');
const wasToolCalled = await rig.waitForToolCall(
'save_memory',
undefined,
(args) => {
try {
const params = JSON.parse(args);
return params.scope === 'project';
} catch {
return false;
}
},
);
expect(
wasToolCalled,
'save_memory should not be called for workspace-specific information',
).toBe(false);
'Expected save_memory to be called with scope="project" for workspace-specific information',
).toBe(true);

assertModelHasOutput(result);
},
Expand Down Expand Up @@ -317,13 +342,13 @@ describe('save_memory', () => {
'Please save any persistent preferences or facts about me from our conversation to memory.',
assert: async (rig, result) => {
const wasToolCalled = await rig.waitForToolCall(
'save_memory',
'invoke_agent',
undefined,
(args) => /vitest/i.test(args),
(args) => /save_memory/i.test(args) && /vitest/i.test(args),

@gemini-code-assist gemini-code-assist Bot Apr 15, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The current predicate uses a loose regex check on the entire stringified arguments of the invoke_agent tool. This can lead to false positives if the agent name is different but the request string happens to contain 'save_memory'. As per repository rules, toolRequest.args is a JSON string and must be parsed using JSON.parse() before its properties can be accessed. It is safer to parse the JSON and explicitly check the agent_name property.

        (args) => {
          try {
            const parsed = JSON.parse(args);
            return parsed.agent_name === 'save_memory' && /vitest/i.test(args);
          } catch {
            return false;
          }
        },
References
  1. The toolRequest.args property is a JSON string, not an object. It must be parsed using JSON.parse() before its properties can be accessed.

);
expect(
wasToolCalled,
'Expected save_memory to be called with the Vitest preference from the conversation history',
'Expected invoke_agent to be called with save_memory agent and the Vitest preference from the conversation history',
).toBe(true);

assertModelHasOutput(result);
Expand Down Expand Up @@ -379,8 +404,15 @@ describe('save_memory', () => {
],
prompt: 'Please save the preferences I mentioned earlier to memory.',
assert: async (rig, result) => {
const wasToolCalled = await rig.waitForToolCall('save_memory');
expect(wasToolCalled, 'Expected save_memory to be called').toBe(true);
const wasToolCalled = await rig.waitForToolCall(
'invoke_agent',
undefined,
(args) => /save_memory/i.test(args),

@gemini-code-assist gemini-code-assist Bot Apr 15, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

Similar to the previous test, using a simple regex on the raw arguments string for invoke_agent is imprecise. Following the repository rule that toolRequest.args is a JSON string, it should be explicitly parsed to verify that the agent_name is save_memory to avoid false positives from other subagent calls.

        (args) => {
          try {
            const parsed = JSON.parse(args);
            return parsed.agent_name === 'save_memory';
          } catch {
            return false;
          }
        },
References
  1. The toolRequest.args property is a JSON string, not an object. It must be parsed using JSON.parse() before its properties can be accessed.

);
expect(
wasToolCalled,
'Expected invoke_agent to be called with save_memory agent',
).toBe(true);

assertModelHasOutput(result);
},
Expand Down
Loading