← All features
Security scanning with a gate
A project-level security scan that treats findings like any other change: evidence first, then a human decision, then a governed fix.
How it works
01
Sweep the repo
Dependency audit, SAST, and secret scanning run against a fresh clone — tool findings and an LLM review pass, deduplicated and severity-ranked.
02
Open a security gate
Findings arrive at a dedicated gate with severity, source, and file references — approve, request changes, or reject.
03
Remediate through the loop
Selected findings become a remediation work item; the coding agent proposes fixes that pass the same proposal gate as feature work, then the project is re-scanned.
The agents behind it
Human control
A dedicated security gate (lead-approved) governs findings; fixes ship only through an approved, cross-checked proposal.