Security Policy

Supported versions

The project is pre-1.0. Until v1.0.0 ships, only the latest minor receives security updates.

Reporting a vulnerability

Do not open a public issue for security vulnerabilities.

Report privately by emailing johnny.iv.young@gmail.com with the subject line [security] electron-stagewright.

Please include:

• A description of the vulnerability and its potential impact.
• Steps to reproduce, or proof-of-concept code.
• Affected version(s).
• Your suggested remediation, if any.

You'll receive an acknowledgment within 5 business days. We aim to provide a disclosure timeline within 14 days of acknowledgment.

Disclosure policy

We follow coordinated disclosure:

1. Confirm the vulnerability and assess severity.
2. Develop a fix in a private fork or branch.
3. Coordinate a release date with the reporter.
4. Publish a security advisory via GitHub Security Advisories (with CVE if applicable).
5. Credit the reporter in the advisory unless they prefer to remain anonymous.

Threat model

The full threat model — assets, trust boundaries, threats and their mitigations, and the residual risks — is published at docs/guides/security-model.md, and the overall posture is recorded in ADR-014.

In one line: the server is a privileged local tool, not a sandbox. It runs with your OS privileges and, under --allow-eval, executes arbitrary JavaScript in the app under test, so only a trusted agent host should invoke it — over the default local stdio transport. Key concerns:

• electron_eval_main / electron_eval_renderer allow arbitrary JS in the app. They are default-deny (unregistered without --allow-eval) with a keyword blocklist and a result cap; the blocklist is defence-in-depth, not a complete control.
• The production plugin reads signed .app bundles and updater feeds. It may return bounded local evidence such as a signing authority in the tool result, but it does not upload that data anywhere.
• The trace plugin writes session artifacts to disk, and the ipc plugin returns captured channel payloads to the agent; configure redact for structured argument/payload fields, and treat screenshots, console output, and tool results as sensitive.

See the threat model for the full analysis and a deployment checklist.