Spec Linter Lite
Drop a spec → see exactly what an AI would do with it. 3-layer validator: format, autotest, rubric.
Lint a specSimulator
Upload an existing AI folder — get the 3 most critical issues, ranked, with fix suggestions.
Free · 12 runs/hour · no email needed
Six months ago you copy-pasted a rules folder to get the agent working. It 'kind of works' now — but it contradicts itself, nobody remembers what's in it, and your gut says don't ship it to prod. Two files quietly tell the agent opposite things. A safety rail is missing that won't show in a happy-path test but will the moment a user types something weird. You can feel the rot; you just can't find the line it lives on.
After completing this
You become someone who can audit a folder you didn't write in under five minutes. Upload it, get the three most critical issues ranked, each with the exact lines quoted and a fix suggested. You learn to tell cosmetic mess from load-bearing rot, and to point at the line that will break first under real traffic — then hand a clean folder to a new teammate without a walkthrough.
You don't read about folder hygiene — you upload a real folder and watch it get sniffed, with the failing lines quoted back at you.
You bring your own (or a sample) folder and react to the ranked issues first; the why-it-matters lands after you've seen the break.
Once you can audit a folder, the paid 'I'm Agent Creator' track teaches you to build folders that pass the sniff by design.
Get a five-minute health check that surfaces the duplicate roles and dead refs you've been stepping around for months.
See the contradictions quoted and line-numbered, so 'it's flaky' becomes 'these two files disagree on line 40.'
Catch missing safety rails before the model swap exposes them — the failures that only show up on weird input.
Not for you if
Folder Sniff pulls the duplicate role definitions and shows you the two files telling it opposite things — quoted, line-numbered, fixable.
Three ranked issues, with quotes. No subjective code review — just what's broken and why it would bite in production.
Missing safety rails get flagged first — the things that wouldn't show up in a happy-path test but will the moment a user types something weird.
Tested on 14 production agent folders before this opened.
Drop a spec → see exactly what an AI would do with it. 3-layer validator: format, autotest, rubric.
Lint a specBuild agent folders end-to-end: pipeline, sub-agents, memory layers, conflict detection.
Try Lite freeMake the numbers honest. Interrogate metrics, audit triage agents, write reconciliation tests, brief AI that surfaces bad news instead of burying it.
Start Ch1 freeYes, fully. Sniff as many folders as you want — no card, no signup.
No. Upload a folder, get the ranked issues. There's no payment step.
A folder sniffs in under five minutes, most of it spent reading the quoted fixes. No prep needed.
Once you can audit a folder, the paid 'I'm Agent Creator' track teaches you to build folders that pass the sniff by design — pipelines, sub-agents, conflict detection.
No. It reads your folder and reports — quotes, ranks, suggests. There's a one-click handoff to the FolderAI editor only if you choose to act on a fix.
Made by mihalkevich.com — runs on the FolderAIThe engine that powers the courses and simulators engine.