OpenSecBench is a security-assessment tool that handles sensitive data — source code, findings, credentials in its vault, and traffic captures. We take the security of the tool itself seriously and appreciate reports made responsibly.

Please do not report security vulnerabilities through public GitHub issues, pull requests, or discussions.

Instead, report it privately through GitHub's private vulnerability reporting: go to the repository's Security tab → Report a vulnerability.

Please include, as far as you can:

• A description of the issue and its impact.
• Steps to reproduce, or a proof of concept.
• Affected version / commit and your environment (OS, LLM provider, runner setup).
• Any suggested remediation.

This is an early-access project maintained on a best-effort basis. We aim to:

• Acknowledge your report within 5 business days.
• Give an initial assessment and a rough timeline within 10 business days.
• Keep you updated as we work on a fix, and credit you when the fix ships (unless you prefer to stay anonymous).

Please give us a reasonable window to address the issue before any public disclosure.

In scope — vulnerabilities in OpenSecBench itself, for example:

• Escapes from the OCI-sandboxed capability/agent runners, or bypasses of the scope guard.
• Leakage of secrets from the encrypted vault, or bypasses of DLP redaction / egress controls.
• Tampering with the append-only, hash-chained audit trail.
• Signature or digest-pinning bypasses in the extension loader / community hub.
• Auth, path-traversal, or injection issues in the local control-plane API.

The security model for these areas is documented in the ADRs under docs/ — notably adr-0011 (secrets/DLP/redaction), adr-0003/adr-0013 (extension trust), adr-0004 (runner protocol), and adr-0002 (provenance/audit). Grounding a report in the relevant ADR helps us triage quickly.

Out of scope:

• Findings that OpenSecBench reports about a target you scan — those are the tool working as intended, not vulnerabilities in OpenSecBench.
• Issues in third-party scanners, models, or extensions — report those upstream.
• Anything requiring a malicious extension the user has explicitly chosen to trust and install (trust is an explicit, signed decision by design), unless it demonstrates a bypass of the stated trust controls.

OpenSecBench is pre-1.0 and moving fast. Security fixes land on the latest main; there is no back-porting to older commits during early access. Always run the latest version.