troubleshoot-conversions

Troubleshoot Conversions

This skill investigates conversion upload issues and generates a structured diagnostic report by executing the mandatory conversion troubleshooting workflow.

1. Execution Instructions [MANDATORY]

When invoked to troubleshoot conversions:

1. Locate the required customer_id (from context or customer_id.txt). If missing, prompt the user.
2. Execute the mandatory diagnostic collector script within the sequestered virtual environment. As it runs, it will output high-level client and action summaries to stdout to inform your analysis:

./.venv/bin/python3 .agents/skills/troubleshoot_conversions/scripts/troubleshoot_conversions.py --customer_id <customer_id> --api_version <api_version>

3. Note the consolidated troubleshooting report path returned by the script (e.g., saved/data/conversion_troubleshooting_report_<epoch>.txt).

2. Diagnostic Queries & Calculations

When analyzing conversion data directly:

• Diagnostic Summaries: Query both offline_conversion_upload_client_summary and offline_conversion_upload_conversion_action_summary.
• Attributes & Totals: Use successful_count and failed_count. Calculate daily total as successful_count + failed_count + pending_count (top-level total_event_count is only available on parent resources).
• Alert Inspection: Access alerts (OfflineConversionAlert) at the top-level resource. Inspect alert.error oneof via WhichOneof("error_code") and report error_percentage. If summaries are empty, append exactly: Reason: No standard offline imports detected in last 90 days.

3. Pre-Upload CSV Validation

Before uploading any conversion CSV data file, execute the programmatic validation utility to statically check formatting and timestamp rules:

./.venv/bin/python3 .agents/skills/troubleshoot_conversions/scripts/validate_conversion_upload.py --csv_path <path_to_csv> --api_version <api_version>