agent-guard is security infrastructure — it enforces policy, isolates credentials and records a tamper-evident audit ledger. A flaw here is a flaw in someone's control plane, so reports are taken seriously and triaged quickly.

Please do not open a public issue for a security bug.

Use GitHub's private vulnerability reporting, which is enabled on this repository:

1. Go to the Security tab
2. Click Report a vulnerability
3. Describe the issue, affected version, and impact

This creates a private advisory visible only to you and the maintainer.

• Affected component (policy engine, DLP, credential broker, audit ledger, SDK)
• Version or commit SHA
• Reproduction steps, ideally a minimal proof of concept
• Impact — what an attacker gains, and what access they need to start

These are targets for a project maintained by one person, not a contractual SLA.

In scope — policy bypass (RBAC/ABAC), DLP exfiltration bypass, credential leakage from the broker, audit-ledger tampering or forgery, approval-flow bypass, privilege escalation between agents or tenants, and authentication or session flaws in the API.

Out of scope — findings that require an already-compromised host or maintainer-level credentials; denial of service from unbounded self-inflicted load; vulnerabilities in third-party dependencies without a demonstrated exploit path through this project; and results from automated scanners submitted without a working proof of concept.

Coordinated disclosure. Report privately, and once a fix ships a GitHub Security Advisory will be published crediting you unless you prefer otherwise. Please give a reasonable window before public discussion.

Good-faith research under this policy is welcome, and no legal action will be pursued for it. Test only against your own deployment — never against another party's data or infrastructure.