ASKP is security-focused infrastructure: it brokers access to AI providers so that applications never hold raw provider credentials. We take the security of the project, and of the people who run it, seriously.

ASKP is in pre-alpha (design phase). There is no released, supported version yet. Once we publish releases, this table will list which versions receive security fixes.

Please do not report security issues through public GitHub issues, discussions, or pull requests. Public disclosure before a fix is available puts users at risk.

Instead, report privately through either of:

1. GitHub private advisory — go to the repository's Security → Advisories → Report a vulnerability page (GitHub Private Vulnerability Reporting). This is preferred.
2. Email — umesh836@gmail.com with the subject line ASKP SECURITY.

Please include, as far as you can:

• A description of the issue and the component affected (e.g. Issuer, Gateway, Vault).
• Steps to reproduce, or a proof of concept.
• The impact you believe it has (e.g. credential exposure, token forgery, tenant crossover).
• Any suggested remediation.

• Acknowledgement within 3 business days.
• An initial assessment and severity rating within 10 business days.
• We will keep you informed of progress and coordinate a disclosure timeline with you.
• With your consent, we will credit you in the advisory and release notes.

We follow coordinated disclosure: we ask that you give us a reasonable window to release a fix before any public discussion. We will not pursue or support legal action against researchers who act in good faith and in line with this policy.

In scope: the ASKP protocol specification and the reference implementation in this repository (Issuer, Gateway, Vault, Policy Engine, Admin API, SDKs).

Out of scope: vulnerabilities in upstream AI providers, third-party dependencies (report those upstream), and issues that require a misconfigured or already-compromised host.

The protocol's core security guarantees are written into the specification as invariants — for example, provider credentials are never returned to clients, policy and budget are enforced before a request reaches a provider, and tokens can be revoked instantly. See spec/askp-protocol-v1.md §3.2 and §9. The full Security Model and Threat Model are forthcoming design deliverables (Batch 2).