Skip to main content
A file scan checks the file you are working on and gives it a health score from 0 to 100. Findings explain what needs attention, such as a secret in the code, repeated code or a function that is difficult to follow. Use the findings to choose what to review rather than treating the score as a guarantee that the code works. Iris Code scans supported files when you open or save them. Open the sidebar’s File tab to see the result for the current file.

The File tab

The tab starts with a short summary: the score, any blockers, which are the most serious findings, whether a Git hook is installed to check code before it is pushed, and which way the score is moving. Below that is the full breakdown:

TypeScript-specific checks

Four extra checks appear for .ts, .tsx, .js and .jsx files. Each one points to a place where TypeScript’s type checking has been bypassed or left incomplete. They are hidden entirely in Go, Python, Ruby, C#, Java, and Rust rather than shown as empty rows.

Code smells

Every finding is clickable and jumps to the exact line.
Turning a toggle off hides both the counter and the detail section, rather than showing a zero. This keeps the sidebar free of categories that are not relevant to your project.

Security smells

Security smells are coding patterns that often lead to a security problem, such as building a database query by joining text together. Iris Code looks for nine of them in every supported language. They count as Blockers, the same category as hardcoded secrets, rather than as warnings. All nine share one toggle, enableSecuritySmells.
To keep them scored without squiggles, leave enableSecuritySmells on and set inlineDiagnostics.securitySmells to false. Full pattern reference in Security Smells.

Warnings

Warnings flag code that has grown too large or is set up unusually. There are five types. Each has a threshold, the limit you set in .irisconfig.json, and a severity you can change on its own.
severityOverrides adjusts any of these. Promote no-exports to a warning where it matters on your project, or demote file-too-long from error to warning while working through a backlog.

How length is counted

A line limit is easy to meet without making code any shorter: join several statements onto one line. Coding agents do this when a file or function is just over its limit, and the result is harder to read and review. So file-too-long, function-too-long and the gate’s gateMaxFileLength cap count every statement after the first on a line as one more line. a(); b(); c() written on one line measures as three lines, the same as writing it on three. When the count includes joined statements, the finding says how many, for example “File is 41 lines, 20 of them statements joined onto shared lines”. Some things that make a line long are not joined statements, and they do not add to the count:
  • A long string, URL or comment.
  • A single long statement, such as a call with many arguments.
  • The parts of a for (...) loop header, and members of an inline type such as { rows: Row[]; total: number }.
  • Short inline markup such as <p>Text with <b>bold</b></p>. Several JSX elements count separately only on a line longer than 100 characters.
Splitting the code into smaller functions or files is what reduces a measured length.
Unused import detection is available for TypeScript, JavaScript and Java. Other languages are not checked for unused imports by Iris Code; their compilers or linters may offer additional checks.Some imports matter even when the name is never written again. Rust imports can make a trait’s methods available, while Ruby and C# can resolve names indirectly at runtime. Removing those imports needs language-specific review.In Java, references in Javadoc links count as usage, so a name used only in documentation is kept.
The refactor playbook turns size and complexity failures into a deterministic plan, mechanical fixes and verification.