agentic-eval

Patterns and techniques for evaluating and improving AI agent outputs. Use this skill when: - Implementing self-critique and reflection loops - Building evaluator-optimizer pipelines for quality-critical generation - Creating test-driven code refinement workflows - Designing rubric-based or LLM-as-judge evaluation systems - Adding iterative improvement to agent outputs (code, reports, analysis) - Measuring and...

Publisher type
Official
Repository
github/awesome-copilot
Skill
agentic-eval
GitHub repository stars
38,956
npx skills add https://github.com/github/awesome-copilot --skill agentic-eval

Agent skills can contain instructions that call tools or modify files. Review the source and permissions before installation.

What this resource does

agentic-eval is a reusable AI-agent instruction package published by GitHub. It gives a compatible agent a repeatable procedure for work in this area: Development.

Best for

  • Founders and teams using an AI coding or work agent for tasks in this area: Development.
  • Repeatable work that matches this capability: Patterns and techniques for evaluating and improving AI agent outputs. Use this skill when: - Implementing self-critique and reflection loops - Building evaluator-optimizer pipelines for quality-critical generation - Creating test-driven code refinement workflows - Designing rubric-based or LLM-as-judge evaluation systems - Adding iterative improvement to agent outputs (code, reports, analysis) - Measuring and...
  • Users who want a publisher-defined workflow they can inspect before running.

What to prepare

  • A compatible AI agent that supports Agent Skills.
  • Access to the project or files the task will use.
  • Any provider accounts, command-line tools, or credentials requested by the skill.
  • A safe branch or backup when the skill may modify files.

How to use it

  1. Open the original repository and review SKILL.md, referenced files, scripts, and requested permissions.
  2. Copy the install command and run it from the workspace where you want to use agentic-eval.
  3. Reload or restart the agent if it does not discover the new skill immediately.
  4. Give the agent a concrete task in this area: Development. Explicitly name agentic-eval when you want to require its use.
  5. Review the proposed actions and output before approving file changes, shell commands, credentials, or external writes.

Example request

Use the agentic-eval skill for this task. First inspect the available context, explain the proposed approach, and wait for approval before any sensitive or destructive action.

Before you continue: Publisher category: Official. Publication in this catalog is not a security guarantee. Review the current repository before installation.

Source and review

Source check: 2026-09-05.

GitHub