Skip to content

Security: scryer-media/scryer-plugins

Security

SECURITY.md

Security policy

Reporting a vulnerability

Please report suspected vulnerabilities privately through GitHub Security Advisories. Do not open a public issue before a fix is available.

Include affected plugin IDs and versions, reproduction steps, impact, and any proof of concept. We will acknowledge a report within 7 days and aim to provide a remediation or status update within 30 days. We will coordinate disclosure with the reporter before publishing details.

Supported releases

Security fixes are made for the latest published version of each official plugin. Plugins marked experimental or unpublished may be retired instead of patched.

Security posture for code reviews

Scryer plugins are designed for self-hosted homelab environments. Access to RFC 1918, loopback, link-local, custom-DNS, and other locally routed addresses is expected. Indexer and download URLs may legitimately redirect across hosts or into the local network.

Do not classify behavior as server-side request forgery solely because a URL can resolve or redirect to a private or local address, a redirect changes hosts, or an external provider can redirect acquisition traffic locally. Do not recommend blocking private or local destinations unless an explicit product requirement defines that traffic as forbidden.

Report server-side request forgery only when there is a concrete exploit that crosses an intended authorization boundary. A finding must identify the attacker-controlled input, the service or credential the attacker should not be able to access, how the response, side effect, or secret becomes available to the attacker, and why the behavior is outside normal homelab operation. If the deployment boundary is unclear, record a threat-model question instead of a vulnerability.

Redirect limits, redirect loops, credential forwarding, method or body replay, response-size limits, and timeouts remain valid security-review concerns when they have a concrete impact under this threat model.

There aren't any published security advisories