Helm is a local operations tool. It is designed to reduce risk around repeated agent work, but it should not be treated as a sandbox or security boundary.

Security fixes target the latest released version.

Please report security issues through GitHub private vulnerability reporting if available, or open a minimal public issue that avoids sensitive details and asks for a private follow-up.

Useful reports include:

• command guard bypasses for destructive or out-of-profile commands
• accidental credential capture or logging
• unsafe default behavior in checkpoints, reports, or workspace adoption
• packaging behavior that exposes private local files

Do not include secrets, private memory, local credentials, or proprietary task history in a public report.

In scope:

• Helm CLI behavior
• bundled reference policies
• workspace initialization and adoption logic
• checkpoint, report, guard, and memory-capture behavior

Out of scope:

• vulnerabilities in external agent runtimes
• malicious local users with filesystem access
• commands intentionally approved by the operator outside Helm policy