Agentlas
Agentlas
← Back to app
Improve

Audit (find ways to improve)

Open any imported repo or any draft and tap the Audit tab. We scan for things that would make your worker less reliable, more dangerous, or harder to use.

What we look for

Skill missing description
It won't auto-trigger. Major issue.
Skill description too short
Less than 60 characters. Triggers will be flaky.
Agent missing tools list
It inherits everything. Usually broader than you wanted.
Hook with destructive shell
rm -rf or sudo without an ask permission. Flagged major.
Memory file too big
Above 50 KB. Will burn tokens on every session.
Duplicate names
Two skills or agents with the same name confuse the router.
No top-level docs
Missing CLAUDE.md / AGENTS.md. Sessions start blind.

What we don't do (yet)

  • Run live third-party AI runtimes for you.
  • Check dependencies for vulnerabilities.
  • Look at your secrets (we hide them, we don't audit them).
HEADS UP · Audit is heuristic, not certified

Each finding has a suggestion you can act on. None of this is a replacement for a real security review of code that handles money or PII.