Agentic Evaluation

Use when measuring or improving an Agent's output with explicit criteria, reflection, evaluator-optimizer loops, rubric or judge comparisons, code-test refinement, and bounded convergence checks.

9 stars 3 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add sandbaseai/workbuddy-skill --skill agentic-evaluation
Manual install
git clone --depth 1 https://github.com/sandbaseai/workbuddy-skill.git /tmp/workbuddy-skill
mkdir -p ~/.claude/skills
cp -r /tmp/workbuddy-skill/skills/agentic-evaluation ~/.claude/skills/agentic-evaluation

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Agentic Evaluation

## Purpose and boundary

Evaluate and improve Agent outputs through a bounded, evidence-based loop:

```text
Generate → Evaluate → Critique → Refine → Verify
```

This is an evaluation workflow, not permission to run arbitrary tools, expose private data, or
accept a model's self-rating as proof. Prefer deterministic checks and independent evaluators
when available. Keep prompts, outputs, traces, and test data within the authorized audience;
redact secrets and personal data before storing or sending them to a judge.

## Evaluation contract

Define before generating or judging:

- **task and expected outcome:** what the output must accomplish and what counts as failure;
- **dimensions and weights:** accuracy, completeness, safety, clarity, format, latency, cost, or other relevant measures;
- **evidence source:** deterministic test, reference answer, human review, retrieval citation, judge, or a combination;
- **threshold and stop rule:** minimum acceptable score, maximum iterations, time/cost budget, and what happens when evidence conflicts;
- **scope and permissions:** models, tools, datasets, environments, retention, and who may approve a release;
- **report contract:** scores, failed dimensions, trajectory, uncertainty, limitations, and follow-up owner.

Separate observed test results from judge interpretation and from the final decision. A score is
not evidence of correctness unless its measurement method and coverage are stated.
Read the full skill on GitHub

Details

Skill name
agentic-evaluation
Path
skills/agentic-evaluation/SKILL.md
License
MIT
Repo updated
8 September 2026

FAQ

What does the Agentic Evaluation skill do?

Use when measuring or improving an Agent's output with explicit criteria, reflection, evaluator-optimizer loops, rubric or judge comparisons, code-test refinement, and bounded convergence checks.

How do I install the Agentic Evaluation skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add sandbaseai/workbuddy-skill --skill agentic-evaluation. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Agentic Evaluation popular?

It lives in sandbaseai/workbuddy-skill, which has 9 GitHub stars and 3 forks.

Related skills

Access Control Security Best Practices

9

sandbaseai/workbuddy-skill

Design, review, implement, and verify least-privilege access controls across users, services, tenants, resources, and administrative workflows.

Security

Accessibility Review

9

sandbaseai/workbuddy-skill

Use when reviewing a web interface or interaction for semantic, keyboard, focus, screen-reader, ARIA, and contrast accessibility risks.

Frontend & Design

Acquire Codebase Knowledge

9

sandbaseai/workbuddy-skill

Use when the user asks to map, document, or onboard into an existing codebase and needs evidence-backed architecture, structure, conventions, integrations, testing, and concerns documentation.

Software Development

AI Governance

9

sandbaseai/workbuddy-skill

Use when classifying AI-system risk, establishing responsible-AI governance, preparing model or system documentation, assessing human oversight, or mapping controls to frameworks such as the EU AI Act and NIST AI RMF.

AI, LLMs & Agents

Agent Supply Chain

9

sandbaseai/workbuddy-skill

Use when generating or verifying integrity manifests, auditing dependency pinning, detecting tampered agent plugins or tools, or establishing provenance gates for agent-component promotion.

AI, LLMs & Agents

AI Prompt Engineering Safety Review

9

sandbaseai/workbuddy-skill

Comprehensive AI prompt engineering safety review and improvement prompt. Analyzes prompts for safety, bias, security vulnerabilities, and effectiveness while providing detailed improvement recommendations with extensive frameworks, testing methodologies, and educational content.

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.