Show HN: Claude Scrub - redact secrets from Claude Code history on disk (github.com)

A new tool called Claude Scrub has been introduced to enhance privacy and security within the Claude model's code history. This tool identifies and redacts sensitive information, such as passwords and API keys, from various historical data maintained by Claude. It systematically replaces secrets with labels, allowing users to manage potentially sensitive information without the risk of exposing it inadvertently in session transcripts, prompt history, and cached inputs. The significance of Claude Scrub for the AI/ML community lies in its ability to automate the redaction process, thereby reducing human error and increasing confidence in using models that might otherwise retain user secrets. The tool relies on a series of detection rules, checking for common secret formats and patterns, while also allowing flexibility for customization through configuration files. Each redaction process is permanent, erasing sensitive data without backups, and ensures that running sessions remain unaffected by the scrub. As AI applications continue to integrate sensitive information, tools like Claude Scrub are essential to safeguard user data, making it a crucial development for developers and practitioners in the AI/ML space.