What problem does it solve?
Sharing company project documents with AI models risks exposing sensitive business information such as client company names, contract amounts, and contact persons. This Skill desensitizes local documents with non-reversible numbered placeholders before any AI analysis takes place.
Core Features & Use Cases
- Regex-Based Masking: Detects Chinese and English company names, project amounts, and contextual person names using a configurable JSON policy.
- Non-Reversible Placeholders: Replaces matched values with numbered placeholders like [mask_1] without creating any restore mapping, and never outputs original matched values.
- Custom Policies: Generate the default policy with generate-policy, save it as JSON, and load custom policies for mask-file runs.
- Use Case: Before asking an AI to analyze project.txt containing client names and contract amounts, run the mask-file command and analyze only the returned sanitized_content or the separate output file.
Quick Start
Ask the AI to use ClawVault to check and mask the local file project.txt first, then analyze only the sanitized content it returns.