SecPriv — Unified Security and Privacy Code Review Skill

Automating security and privacy code review involves using an LLM agent to analyze pull requests for CWE-mapped weaknesses and GDPR-mapped violations. It returns a structured JSON array of high-confidence findings for Python and JavaScript/TypeScript.

A unified detector-validator workflow enumerates candidate findings across security and privacy surfaces, then suppresses false positives using a shared, confidence-thresholded validator. It enforces a five-phase process mapping source/sink flows to CWE or GDPR standards.

Yes, you can check Python and JavaScript/TypeScript code for GDPR compliance in a single pass. The review performs dual-surface assessment to identify privacy violations mapped to GDPR articles alongside CWE security weaknesses.

Yes, automated LLM code review maps findings to specific CWE IDs. It requires high-confidence findings to include explicit CWE references for security weaknesses or GDPR articles for privacy violations in the output.

Reduce false positives in automated security code review by applying validator suppression rules after initial source/sink mapping. The shared validator thresholds confidence levels before returning the final JSON array of findings.

Limitations include support for only Python and JavaScript/TypeScript code. It requires high-confidence findings mapped to GDPR articles and is intended for reviewing single files, diffs, or pull requests.