METRMETROfficialยท2 Agent Skills Included

hawk

Run and debug AI evaluations at scale on AWS

Runs Inspect AI evaluations and Scout scans as isolated Kubernetes jobs on AWS from a single YAML config. Removes manual pod provisioning, API key handling, and log collection with a managed LLM proxy and PostgreSQL results warehouse. Helps engineers monitor live runs, debug stuck evaluations, audit dependency bumps, and manage database migrations faster.
npx skills add METR/hawk --all -g -y
Available:

Gives the agent the project overview, coding standards, secrets and observability rules, Pulumi deployment workflow, and pointers to nested per-directory instruction files so it works on Hawk correctly.

All Skills in This Repository (2)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install Hawk?โ–ผ

Run `npx skills add METR/hawk --all -g -y` in your terminal to install all skills in this suite globally.

What does Hawk do?โ–ผ

Hawk runs Inspect AI evaluations and Scout scans on AWS, provisioning isolated Kubernetes pods, proxying LLM API access, and storing results in a PostgreSQL warehouse with a web UI.

How do I debug a stuck evaluation?โ–ผ

Use the debug-stuck-eval skill: start with `hawk watch <eval-set-id>` for a live snapshot, then check traces, stack traces, and logs to find where the runner is blocked.

Can I view and export evaluation results?โ–ผ

Yes. The view-results skill uses the hawk CLI to list eval sets, inspect samples and scores, and download full transcripts individually or in bulk.

Do I need deep AWS or Kubernetes experience to use these skills?โ–ผ

No. The skills wrap the hawk CLI and ops scripts in plain-English workflows, so your agent handles the underlying infrastructure commands for you.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’