view-results

List, inspect, and retrieve detailed results from Hawk AI evaluations.

25|11|Updated Feb 6, 2025
One-click install
npx skills add https://github.com/METR/inspect-action --skill view-results-metr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: view-results
Source: https://github.com/METR/inspect-action/tree/main/.claude/skills/view-results
Command: npx skills add https://github.com/METR/inspect-action --skill view-results-metr

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps users understand and analyze the outcomes of Hawk evaluations by providing tools to list, inspect, and retrieve detailed results from completed evaluation runs.

Core Features & Use Cases

  • List Evaluations: Browse and identify specific evaluation sets and individual evaluations.
  • Inspect Samples: View individual sample results, scores, and statuses.
  • Retrieve Transcripts: Download full conversation logs for detailed analysis of agent behavior.
  • Use Case: After running a batch of AI model evaluations, use this Skill to list all completed eval sets, then drill down into a specific eval set to see which models performed best, and finally download the transcripts for the lowest-scoring samples to understand where the models failed.

Quick Start

List all available eval sets by running the command hawk list eval-sets.

Frequently Asked Questions about view-results

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I view Hawk evaluation results for completed runs?

To view Hawk evaluation results, you can list available evaluation sets and inspect individual samples to review scores, statuses, and detailed outcomes from completed AI evaluation runs.

What's the best way to analyze agent decision-making from evaluation transcripts?

You can analyze agent decision-making by retrieving full conversation transcripts from specific evaluation samples, allowing you to review detailed logs and understand where the agent succeeded or failed.

How do I list all available evaluation sets in Hawk?

You can list all available Hawk evaluation sets by running the command `hawk list eval-sets`, which browses and identifies specific evaluation sets and individual evaluations for review.

Do I need the Hawk CLI to retrieve evaluation results and transcripts?

Yes, retrieving detailed evaluation results and transcripts requires access to the Hawk CLI along with configured API environments to successfully fetch data from completed runs.

Can I identify which AI models performed best by inspecting Hawk evaluation samples?

Yes, you can inspect individual sample results and scores within an evaluation set to compare model performance and identify which models performed best or failed during the evaluation.

How do I find failed runs in a batch of Hawk AI evaluations?

You can find failed runs by listing all completed evaluation sets, drilling down into a specific set to inspect sample statuses, and identifying which samples recorded failures or low scores.