Help
How does the scan work?
Three layers: pattern rules catch known mistakes instantly, a code-structure check (AST) finds slow loops, ignored errors and tangled functions, and optional AI review (Groq) reads the code for deeper problems.
Why are test files skipped?
Test suites and fixtures contain intentionally broken or vulnerable sample code, so scanning them mostly reports demo strings rather than real problems. Build output, minified bundles and test files are skipped automatically; real leaked secrets are best caught before commit with a tool like gitleaks (see the tips under any scan).
Is my code private?
Files stay in your browser. Scanning sends them to this app's server function, which runs the checks and forgets them. If you add a Groq key, code snippets are also sent to Groq for AI review.
What does the health score mean?
100 is clean. Issues subtract points by severity (critical 30, high 12, medium 5, low issues capped at 10 total), averaged over the files that have issues and run through a soft curve — so problems spread thinly across a big repo don't zero the score, while genuinely bad files still do. Fixed and ignored issues don't count.
What does 'Fix all' change?
It rewrites only lines with a safe one-line fix and gives you a ZIP of the patched files. Always review and test before deploying. Some fixes are placeholders, for example your real domain for CORS.
Why is AI review partial?
The free AI tier has limits. We review the first 8 files, and pause if Groq says it's busy. Rule-based checks always cover everything.
How do share links work?
The report is packed into the link itself, so nothing is stored on a server. Anyone with the link can read it until it expires. Source code lines are not included.
Keyboard shortcuts
On results: / jumps to the filter, e exports the Markdown report.