How can I save a full status report to a file?

Run 'hawk status > report.json' to export the complete job status and metrics into a JSON file for further analysis.

Can I stream live logs using this skill?

No, the -f or --follow flag should not be used as it blocks the interface. Instead, run 'hawk logs' periodically to see updated output.

What is a JOB_ID in the context of Hawk monitoring?

The JOB_ID refers to the unique eval_set_id or scan_run_id assigned when a job is submitted to the Inspect-action cloud environment.

How do I check logs for my most recent Hawk job?

You can use the 'hawk logs' command without a job ID to automatically retrieve the last 100 log lines for the most recent evaluation or scan run.

Hawk Job Monitoring

Name: Hawk Job Monitoring
Author: METR

byMETR

•

分析与监控

Monitors Hawk job status, retrieves logs, and diagnoses issues for AI evaluation runs within the UK AISI Inspect framework.

关于

The Hawk Monitoring skill allows developers to track and manage evaluation jobs and scans running in the cloud. It provides seamless access to the Hawk CLI's logging and status reporting capabilities, enabling users to check job progress, debug failures via specific log segments, and generate detailed JSON or text reports for comprehensive analysis of Inspect evaluation runs.

主要功能

Diagnose execution errors in remote evaluation runs
Filter log output by specific line counts
Retrieve logs for specific or most recent Hawk jobs
Generate comprehensive status reports for job analysis
12 GitHub stars
Retrieve historical logs within a defined hour window

使用场景

Exporting full job status reports for post-run documentation and review
Debugging failing jobs by extracting and analyzing error logs
Checking the real-time progress of a long-running AI evaluation job

关于

主要功能

Diagnose execution errors in remote evaluation runs
Filter log output by specific line counts
Retrieve logs for specific or most recent Hawk jobs
Generate comprehensive status reports for job analysis
12 GitHub stars
Retrieve historical logs within a defined hour window

使用场景

Exporting full job status reports for post-run documentation and review
Debugging failing jobs by extracting and analyzing error logs
Checking the real-time progress of a long-running AI evaluation job