Why This Matters

failed-tool counting and exit-code logic are product-visible details.

Read First

• app/services/tracer_client/tracer_tools.py

Scope

1. Add a fake subclass that stubs _get().
2. Test get_run_tasks() with all-success, mixed-failure, and empty payloads.
3. Confirm failed-task details include the expected extra fields.

Tests / Coverage

• tests/services/tracer_client/test_tracer_tools.py
• cover non-zero exit code handling
• cover None, empty string, and "0" exit code cases
• cover unsuccessful responses

Acceptance Criteria

• task-count logic is directly protected
• failure-detail extraction is tested
• response handling is fully offline