$ npx agnix .
Validating: .
.claude/settings.json:8:3 error: Hook command contains dangerous pattern 'rm -rf' [CC-HK-009]
help: Hooks run automatically -- destructive commands can cause data loss
.claude/skills/deploy/SKILL.md:1:1 error: Dangerous skill without safety flag [CC-SK-006]
help: Add 'allowedTools' restrictions to prevent unrestricted access
CLAUDE.md:15:1 warning: Generic instruction 'Be helpful and accurate' [fixable]
help: Remove generic instructions. Claude already knows this.
Found 2 errors, 1 warning
1 issue is automatically fixable
hint: Run with --fix to apply fixesSee it in your editor
Try it right now — no install needed
Paste your CLAUDE.md, SKILL.md, or any agent config and see diagnostics instantly. Runs entirely in your browser via WebAssembly.
Open Playground230 Validation Rules
Comprehensive validation across 31 rule categories. Catch broken skills, invalid hooks, misconfigured MCP servers, and much more.
Auto-Fix
Fix common issues automatically. Run agnix --fix . and move on.
Editor Integration
Real-time diagnostics in VS Code, Neovim, JetBrains, and Zed via the built-in LSP server.
What does agnix catch?
Your skill was invisible to the agent. Now it triggers.
Hooks run automatically. This could delete your project.
Claude already knows this. It was wasting your context window.
The agent was parsing your instructions incorrectly.
