How it works
Advisory to runnable proof, autonomously.
Analyze the advisory
An AI agent reads the GHSA or CVE, pulls the vulnerable package, and understands the vulnerability's root cause.
Reproduce in a sandbox
The agent builds a faithful copy of the affected software and fires the exploit, capturing the crash, the leak, or the shell.
Publish verifiable proof
A self-contained script, a session replay, and a permanent REPRO ID. Run pruva-verify to see it fire yourself.
Latest
Verified Reproductions
HashiCorp Nomad's Docker task driver can be tricked into bind-mounting host paths via a symlink containment bypass, enabling sandbox escape and host file read/write.
CodeIgniter4 View Parser conditional tags allow arbitrary PHP code execution in user-editable templates
vm2 before 3.12.1 sandbox escape via host-realm Promise Symbol.species (VM and NodeVM)
gVisor directfs openHandle host-FD identity TOCTOU → guest-to-host sandbox escape (follow-up to CVE-2026-96812, no separate CVE assigned)
IBM Langflow OSS 1.0.0 through 1.12.2 is vulnerable to remote unauthenticated code execution via OS command injection.
Ghost CMS RCE via Bookmark Card Images (SVG handling in bundled image processing library)
The identifier
Why REPRO IDs?
Verified Proof
Each reproduction includes executable scripts, session replays, and before/after evidence proving the vulnerability exists.
Permanent Citation
REPRO IDs are permanent, citable references. Link CVEs, GHSAs, and issues to verified reproductions.
Full Transparency
Watch session replays showing exactly how the agent reproduced the issue. Nothing hidden, everything auditable.
Cite it
Embed in Your README
Show that your security advisory has been independently verified with an embeddable badge.
Don't take the advisory's word for it. Run it.
Browse the full catalog of autonomously reproduced vulnerabilities — each with a runnable proof and a permanent REPRO ID.
Browse All Reproductions →