Install
openclaw skills install know-your-aiCommand-line tool for red-team AI security testing, running evaluations, detecting vulnerabilities, and reviewing detailed results via API integration.
openclaw skills install know-your-ai---
name: know-your-ai
description: AI security testing & evaluation CLI. Run red-team evaluations, check vulnerabilities, and review results for your AI products.
homepage: https://knowyourai.hydrox.ai
requires:
bins:
- node
env:
- KNOW_YOUR_AI_DSN
primaryEnv: KNOW_YOUR_AI_DSN
metadata: {"clawdbot":{"emoji":"🛡️","requires":{"bins":["node"],"env":["KNOW_YOUR_AI_DSN"]},"primaryEnv":"KNOW_YOUR_AI_DSN"}}
---
# Know Your AI
AI security testing and evaluation CLI by [HydroxAI](https://hydrox.ai). Run red-team evaluations, detect jailbreak vulnerabilities, and review security scores for your AI products — all from the command line.
## Check setup
```bash
node {baseDir}/scripts/doctor.mjs
Validates your DSN configuration and tests API connectivity. Requires KNOW_YOUR_AI_DSN environment variable.
node {baseDir}/scripts/target.mjs
node {baseDir}/scripts/list.mjs
node {baseDir}/scripts/evaluate.mjs <evaluation-id>
node {baseDir}/scripts/evaluate.mjs <evaluation-id> --max-prompts 5
node {baseDir}/scripts/evaluate.mjs <evaluation-id> --timeout 300
Triggers an evaluation run and streams real-time progress. Returns scores, pass/fail counts, and the run ID.
--max-prompts <n>: Maximum prompts per dataset (default: 3)--timeout <seconds>: Maximum wait time in seconds (default: 600)--debug: Enable debug loggingnode {baseDir}/scripts/history.mjs
node {baseDir}/scripts/history.mjs --all
--all / -a: Show all runs (default: last 10)node {baseDir}/scripts/result.mjs <run-id>
node {baseDir}/scripts/describe.mjs <evaluation-id>
Shows detailed evaluation configuration: judge model, threshold, linked datasets, and prompt counts.
node (>=18) runtimeKNOW_YOUR_AI_DSN environment variable from the Know Your AI dashboard (Settings → API Keys)https://kya_xxx:da2-xxx@host/product_idkya is also available if installed globally via npmdoctor first to verify connectivity before running evaluations