fix: align E2E batch eval and recommendation tests with current API

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀

[notgitika]

Summary

• batchEvaluateId → batchEvaluationId in assertion (field was renamed in API migration feat: make memory retrieval config conditional on strategies #142)
• Invoke prompt changed from "Say hello" to "What is 3 + 5? Use the add_numbers tool." so the agent generates tool traces in CloudWatch
• Tool recommendation test changed from search/calculator (which the agent doesn't have) to add_numbers (the actual tool in the Strands template)

Root cause

The E2E test was written against the old batch evaluation schema before the API migration in 300a5168. The tool recommendation test used tool names that don't exist in the Strands template agent, causing the API to return a ValidationException because those tools never appear in agent traces.

Test plan

• config-bundle-eval-rec.test.ts batch eval tests should no longer fail on assertion mismatch
• Tool recommendation test should find add_numbers traces from the updated invoke

[agentcore-cli-automation]

Reviewed the changes — all three look correct and well-targeted:

1. batchEvaluateId → batchEvaluationId matches the current API client shape in src/cli/aws/agentcore-batch-evaluation.ts (no other stale references remain).
2. The updated invoke prompt will actually exercise a tool call, which is what the later tool-description recommendation test needs in CloudWatch traces.
3. add_numbers is the tool defined in all Strands template variants (src/assets/python/{agui,a2a,http}/strands/base/main.py), so the recommendation test will now find matching traces instead of hitting ValidationException.

LGTM 🚀