Destructive command refusal
The command guard refuses known destructive shapes before they run and prints the reason.
Security
ForgeOS earns trust by making its limits visible. This page states what the checks enforce, what they do not guarantee, and where to report a vulnerability without making it public.
Cited release v1.20.1
The hooks and validation checks are a safety net against accidents, not a security boundary. They reduce preventable mistakes during governed delivery; they do not replace review, least privilege, or a real threat model.
These checks are mechanical controls for governed delivery. They catch known unsafe patterns and unauthorized writes before a task is treated as complete.
The command guard refuses known destructive shapes before they run and prints the reason.
Discovery and governance guards refuse writes outside authorized paths; implementation windows fail closed when missing, empty, malformed, or too broad.
The secret scanner blocks credential-shaped strings before a change is called complete.
Validation refuses a completion claim unless the relevant checks were executed and observed.
Report vulnerabilities through the repository security advisory route, not as a public issue. Include what you found, the file or script involved, steps to reproduce, and the impact. Do not include a real secret value.
The repository remains the canonical source for policy text, release files, and validation runs. This page cites the public release below.
Release v1.20.1Release notes for v1.20.1 on this siteValidation runsVerify a release download