Evaluate and refine AI agent outputs through structured self-critique loops.
npx skills add https://github.com/ncksol/ravebot --skill agentic-eval-ncksol
Please help me install this Agent Skill. Skill: agentic-eval Source: https://github.com/ncksol/ravebot/tree/main/.github/skills/agentic-eval Command: npx skills add https://github.com/ncksol/ravebot --skill agentic-eval-ncksol
Agents often produce outputs that require rigorous quality control to ensure reliability, consistency, and alignment with goals. This skill provides systematic evaluation and refinement workflows to improve accuracy and trustworthiness across AI tasks.
Run a basic evaluation loop on a sample task to generate and refine the output.
High-intent search queries and answers about installing and using this skill.
LLM-as-judge evaluates AI agent outputs by applying rubric-based scoring within structured evaluation loops. It generates JSON feedback to critique and refine high-stakes tasks like code generation, ensuring outputs align with predefined quality criteria and reliability goals.
Iterative evaluation for code generation is set up by running structured self-critique loops that evaluate outputs against a rubric and provide JSON feedback. The workflow separates generation and evaluation interfaces, applying guardrails to ensure safe convergence toward high-reliability code.
You do not need external dependencies to run iterative evaluation loops, as this skill operates independently. It provides built-in guardrails for convergence and safety, allowing you to apply self-critique workflows to AI tasks without requiring additional libraries or frameworks.
AI agent output lacks consistency when generation and evaluation are not separated into distinct components. Implementing structured self-critique loops with rubric-based scoring addresses this by iteratively refining outputs against strict criteria, ensuring alignment with reliability and consistency goals.
githubAutomate iterative evaluation and refinement of AI agent outputs with JSON feedback.
Owl-ListenerCodify format, length, tone, and content boundaries into reusable constraint specifications.
jaimedhenriquesEvaluate and iteratively refine AI outputs using structured JSON feedback.
forgivesam168Evaluate AI agent outputs using structured rubrics and JSON feedback loops.
