Basalt is a defensive security tool. It finds vulnerabilities, writes proof-of-concept exploits, and verifies them by execution in an isolated sandbox. Only run Basalt against:
- code and infrastructure that you own, or
- targets you have explicit written authorization to test.
Using Basalt against systems you do not own or are not authorized to test may
be illegal. You are solely responsible for how you use this tool. The authors
accept no liability for misuse (see LICENSE).
- Basalt executes generated PoCs and target code inside a Docker sandbox with
networking disabled (
--network none) and resource limits. Do not disable this isolation when auditing untrusted code. - The agents read untrusted target content and are therefore a prompt-injection surface. Basalt mitigates this (content spotlighting, secret redaction, and a design where model agents never invoke tools directly), but treat any host running audits as you would any code-execution environment.
If you discover a security issue in Basalt itself, please report it privately — do not open a public issue.
- Use GitHub's "Report a vulnerability" (Security → Advisories) on this repository, or
- open a minimal private channel with the maintainer.
Please include: affected version/commit, a description, and a reproduction if possible. We aim to acknowledge reports within a few days and will coordinate a fix and disclosure timeline with you.
Basalt is pre-1.0. Security fixes land on main; pin a commit for
reproducibility.