Help › Keys, CLI and automation

Catch untranslatable text before you commit

Partly builtLast updated 2026-10-29

Some text never reaches translation because of how it is written. rl-lint finds the common cases in your source files, so you can fix them before you commit.

It is a rough, line-by-line check. It follows the same rules the product uses to find text. It can miss things and can warn about harmless ones. Warnings do not fail a run unless you ask. The one error is a committed secret key.

What it finds#

• A secret key (rl_sec_) in a file. It is never printed in full.
• Markers the product ignores, such as translate="false" or the class no-translate. The right ones are translate="no" and notranslate.
• Words inside noscript or template, and words in SVG text.
• Images without alt text.
• Words in attributes that are not collected, such as data-tooltip or aria-description.
• A word split by markup, like Hel<span>lo</span>.
• Text set from code, such as textContent = "..." or alert("..."). A build-time step cannot see it.
• A sentence glued from pieces with + or ${}. Translators need the whole sentence.
• Several unnumbered %s or %d in one string.
• Words in CSS content:.
• A page without lang on the html tag.

Silence a finding#

Add rl-lint-disable-next-line RL105 in a comment on the line above, or rl-lint-disable-line. To switch a rule off, pass --ignore with its code.

Where to run it#

By hand on a folder, in CI with GitHub-style output, or as a git pre-commit hook on staged files. Output formats are text, GitHub annotations and JSON.

Status#

It is in the repository and not published: OPEN.