@rhetorlint/cli

Read the subtext from the command line. RhetorLint marks rhetorical tells in text, emits human-readable, JSON, or SARIF output, and can gate CI on tell density. Analysis runs locally with no model calls, network requests, or telemetry.

RhetorLint reads the language, never the person. A mark identifies a visible language pattern; it does not detect lies, infer intent, or judge whether a claim is factually true.

Install

npm install --save-dev @rhetorlint/cli
npx rhetorlint --help

It also works without adding a dependency:

npx @rhetorlint/cli --version

Use

Analyze files or pipe text on standard input:

rhetorlint statement.txt
echo "Mistakes were made." | rhetorlint --json
rhetorlint --sarif press-release.txt > rhetorlint.sarif
rhetorlint --max 8 comms/*.md

--max <n> exits with status 1 when any input exceeds the specified number of tells per 100 words, making it suitable for a CI check. Usage errors exit 2.

| option | effect | |---|---| | --json | emit the canonical RhetorLint JSON result | | --sarif | emit SARIF 2.1.0 for editors and code scanning | | --max <n> | fail when tell density exceeds the threshold | | --rules <path> | use a custom JSON rule pack | | --quiet | suppress the human report | | --no-color | disable ANSI colour | | --version | print the CLI version | | --help | print command help |

The CLI installs @rhetorlint/core and the English rule pack @rhetorlint/rules-en. The specification and source are available in the RhetorLint repository.

MIT.