Evaluate and iteratively refine AI agent outputs through self-critique loops.
npx skills add https://github.com/junevm/cdns --skill agentic-eval-junevm
Please help me install this Agent Skill. Skill: agentic-eval Source: https://github.com/junevm/cdns/tree/main/.github/skills/agentic-eval Command: npx skills add https://github.com/junevm/cdns --skill agentic-eval-junevm
This Skill addresses the challenge of generating high-quality, reliable AI outputs by enabling agents to critically evaluate and iteratively improve their own work.
Use the agentic-eval skill to refine the generated code based on the provided test cases.
High-intent search queries and answers about installing and using this skill.
To implement a self-critique loop for AI-generated code, you use iterative refinement patterns where an agent writes code, evaluates it against test cases, identifies bugs, and automatically fixes them until quality standards are met.
An evaluator-optimizer pipeline is an agentic workflow pattern that generates outputs, evaluates them against defined criteria or rubrics, and iteratively refines the results through self-critique until the desired quality threshold is reached.
Using LLM-as-judge for rubric-based evaluation involves having an AI agent assess its own outputs against structured criteria, requiring structured output formatting to reliably parse critique results and trigger automated refinement steps.
Yes, structured output is required for reliable parsing of critique results during iterative refinement, ensuring the agent can accurately extract identified issues and apply robust error handling for automated improvement loops.
You should use test-driven code refinement in an agentic workflow when generating complex code, allowing the agent to automatically test its output, identify bugs through self-critique, and iteratively fix issues for a robust final product.
htn332805Evaluates and improves AI agent outputs via self-critique and iterative refinement.
pluginagentmarketplaceRun Python-based prompt-injection and adversarial security tests in CI/CD pipelines.
spencerbeggsHarden recursive-descent parsers against stack exhaustion, complexity attacks, and prototype pollution.
MudassarAbrarEvaluate and refine AI agent outputs through iterative generate-evaluate-critique-refine loops.
