Agent Skill

Code review that reads your repo.

Yuno turns your coding agent into a sparring partner — it greps your code, finds what you missed, and challenges you with line numbers.

Paste this into Claude Code, Cursor, or any coding agent:

Read https://yuno.md/SKILL.md and install the skill

Your agent reads the skill, installs it, and activates it automatically.

The difference

> "Review the path validation in primitives.rs" 
// primitives.rs
fn is_within_allowed(path: &str, allowed: &[String]) -> bool {
    let canonical = fs::canonicalize(path).unwrap();  // line 47
    allowed.iter().any(|dir| {
        let allowed_canonical = fs::canonicalize(dir).unwrap();
        canonical.starts_with(allowed_canonical)
    })
}
Your agent
"Looks good. The function correctly checks if a path is within the allowed directories by canonicalizing both paths and comparing them. Implementation looks solid."
Your agent with Yuno
"is_within_allowed re-canonicalizes every path on each call — line 47. That's O(n) syscalls per validation and will silently resolve symlinks, which means a path outside your allowed dirs could pass if there's a symlink inside them. Cache the canonical allowed paths at init, and decide if you want to follow symlinks or not — those are two different security models."

Why Yuno instead of your own prompt?

Because Yuno doesn't just critique — it reads your code first. Any prompt can say "question my assumptions." Yuno greps your repo, finds the actual complexity you didn't mention, and throws it back at you with line numbers.

Compatible with:

Claude Code Cursor Roo Code Amp VS Code Gemini CLI OpenAI Codex