Multi-tier framework for evaluating AI agent skills and plugins with quality gates, security scanning, harness compatibility checks, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior. Ships reusable CI templates that bring every check to any repository.
Python agent-evaluationagent-pluginsagent-securityagent-skillsagentic-aibenchmarkci-templatesclaude-codecodexevaluateevaluationharnesspluginplugin-evaluationsecurity-scannerskill-evalskill-evaluationskill-evaluatorskill-gap-analysisskills
---
title: "Run your first check"
sidebar-title: "Quickstart"
description: "Get a bundled skill, run a quality check, and read the saved result. No API key required."
---
Get a quality score and findings for a complete example skill.
[Install SkillEvaluator](/skills/skillevaluator/installation#install) first,
then return here. The lightweight installation is enough; this check needs no API key, GPU, or Docker.
<Steps toc={true}>
### Get the example
You need Git to download the example.
```bash title="Clone the repository"
git clone --depth 1 https://github.com/NVIDIA/SkillEvaluator.git
cd SkillEvaluator
```
The bundled `text-analyzer` skill includes instructions and a working script.
Keep this terminal open for the following commands.
### Run without an LLM
```bash title="Score the bundled skill"
skillevaluator quality-check \
src/skillevaluator/tier3/reference_skills/text-analyzer \
-r cli -r json -o reports/first-quality
```
The check runs offline with your installed CLI.
### Read your result
The terminal shows your **score out of 100**, **A–F grade**, and findings to fix.
The default minimum quality score is **70**. Open the saved JSON result at:
```text title="Your saved report"
reports/first-quality/skillevaluator-quality.json
```
<Note>
This is a quality check. Security scanners and live agents have not run,
so the report's overall publication status remains **INCOMPLETE** even when
the quality check passes.
</Note>
</Steps>
## Check your own skill
Replace the sample path with the directory containing your `SKILL.md`.
Keep the report output outside that directory:
```bash title="Run from the same checkout"
skillevaluator quality-check ./my-skill \
-r cli -r json -o reports/my-skill-quality
```
## Use NVIDIA Build
Want model feedback on the instructions? Add a rubric review for the same skill.
<Accordion title="Set up a key and run a rubric review">
Add the `llm` extra using [Installation Options](/skills/skillevaluator/installation-options#choosing-extras),
or choose the [all-tier installation](/skills/skillevaluator/installation#all-tiers) first.
1. Sign in to [NVIDIA Build](https://build.nvidia.com/) and complete the
Developer Program enrollment prompts.
2. On a model page, choose **Get API Key** and copy your key. See the
[API Catalog quickstart](https://docs.api.nvidia.com/nim/docs/api-quickstart)
for the account setup steps and the
[access FAQ](https://docs.api.nvidia.com/nim/docs/product) for eligibility and limits.
3. In the same terminal, replace the example key and run:
```bash title="Add model-backed feedback"
export SKILL_EVAL_LLM_PROVIDER=nv_build
export NVIDIA_API_KEY='nvapi-...'
skillevaluator rubric-eval \
src/skillevaluator/tier3/reference_skills/text-analyzer \
-r cli -r json -o reports/first-build
```
Read the rubric feedback in the terminal or open
`reports/first-build/skillevaluator-rubric.json`.
Scores can vary between calls. This reviews the instructions; it does not
run an agent with the skill. No local GPU or Docker is needed.
For other providers or model overrides, see
[Providers & Credentials](/skills/skillevaluator/configuration).
</Accordion>
## Where to go next
<CardGroup cols={2}>
<Card title="Run a live comparison" icon="flask" href="/skills/skillevaluator/tier3-live-evaluation">
Measure whether a skill improves an agent's task results.
</Card>
<Card title="Add automatic checks" icon="code-branch" href="/skills/skillevaluator/repository-integration">
Start with a GitHub or GitLab CI template.
</Card>
<Card title="Run full static validation" icon="shield" href="/skills/skillevaluator/installation-options#system-tools">
Check scanner setup for Semgrep, SkillSpector, and Gitleaks, then run Tier 1.
</Card>
<Card title="Try a guided experiment" icon="terminal" href="/skills/skillevaluator/starter-demo">
Change the example, introduce a regression, and compare real outputs.
</Card>
</CardGroup>
Find overlapping skills with [Tier 2](/skills/skillevaluator/tier2-deduplication),
or learn more about your [reports](/skills/skillevaluator/reports).