Preferred: use GitHub private vulnerability reporting for this repository (Security tab -> Report a vulnerability). If that is not available to you, report privately by email to ckellyreif@gmail.com rather than opening a public issue. Include what you found, how to reproduce it, and the impact you see.

Acknowledgement SLA: within 3 business days. If a report is confirmed as a real vulnerability, expect a fix or mitigation plan communicated within 14 calendar days given the solo-maintainer scale of this project; critical, actively-exploitable findings are prioritized ahead of that window.

Please do not post intake data, exports, or any real personal data in a report. A minimal synthetic reproduction is enough.

This project is pre-1.0. Fixes land on the latest minor release. There is no backport guarantee until 1.0.

The tool reads untrusted files (CSVs, intake PDFs and scans, plain-text and .eml bodies) and writes to a case system. The concerns it takes seriously:

• Parsing untrusted input. Malformed CSVs must fail with a clear error, not a crash that leaks a stack trace with data in it. PDF parsing runs in a resource-limited child process by default (containment, not a full syscall sandbox; see docs/THREAT-MODEL.md).
• Consent enforcement. Under a consent-required policy, a record without granted consent must never appear in an export or in logs. This is enforced by tests, not by review alone.
• Data minimization in artifacts. The withheld.csv file lists ids and a reason, never field values, so the record of what was withheld does not itself leak the data.

Network hardening, multi-tenant isolation, and authentication are out of scope while the tool runs locally against an operator's own data. The DV policy pack keeps inference offline so that PII does not egress; the threat model for the untrusted-document parse path is documented in docs/THREAT-MODEL.md.