English | Русский

This project accepts source text and, when explicitly requested, sends a process description to a configured LLM provider. The core linter, Markdown handling, and offline synthesizer do not require network access or optional model packages.

Report suspected vulnerabilities through the private security-advisory mechanism provided by the repository hosting the project. Do not include API keys, access tokens, private customer data, or unreleased source code in a report. If private reporting is unavailable, contact the maintainers through the channel documented by the hosting organization.

Do not open a public issue containing exploit details or sensitive data.

• Use a dedicated, least-privilege API key for LLM experiments and rotate it after use.
• Review the selected provider's retention, training, and transport policies before sending descriptions.
• Keep .env files out of version control. The repository ignores .env and its variants while retaining .env.example.
• Treat generated Mermaid as untrusted text. A diagram source is not a safe HTML or script policy mechanism.
• Review repairs before writing them into production documentation. The tool preserves uncertain input and reports failures, but users remain responsible for the final source.
• Use --force only when replacing the selected output is intentional. The CLI rejects symbolic-link output paths and uses atomic replacement for regular files.
• Keep input size limits enabled when processing files from untrusted locations.

The latest released version is the supported line. Security fixes are applied to the current release line unless a project release note states otherwise.