anonymize uploaded pdf, powerpoint, and word files with presidio. use when the user provides an uploaded pdf, ppt/pptx, doc/docx, or wants a fully redacted version that removes names, emails, phone numbers, addresses, ids, and company/project names. only use on files uploaded in the current conversation.
Use this skill to turn an uploaded PDF, PowerPoint, or Word file into a fully redacted version. Detect names, emails, phone numbers, addresses, IDs, and company or project names with Presidio, then replace matched content so the resulting file no longer contains the original sensitive text.
Use these entity groups for redaction:
Redact completely rather than substituting placeholders. Prefer visual removal over partial masking.
Use Presidio as the detection layer and redaction engine for text entities. The analyzer finds sensitive spans and the anonymizer applies the redaction operator. Default to a full redaction operator for every target entity.
When needed, extend Presidio with project-specific recognizers for company and project names so they are also removed.
Before returning the output:
Only process files uploaded in the current conversation. Do not use connector files or external sources for this skill.