Dangerous Command Guard
A confirmation step for some risky commands submitted through Chau7.
Availability
Partially implemented in stable. Read the limits below before relying on this workflow.
Reviewed against stable v0.5.1. Rust candidate additions are labeled separately. Release status.
The pain this solves
- A prompt can help catch some risky command submissions.
- An input guard is not shell isolation or universal script protection.
What ships with it
- Stable code classifies command lines and can request confirmation.
- The guard covers supported Chau7 terminal or MCP submission paths.
- It is not a sandbox and does not cover arbitrary scripts or remote execution.
Questions this answers
- Can Chau7 prompt for some risky submitted commands?
- Is the guard a shell sandbox?
- Does it inspect every hidden-character paste and script?
How it works
The stable DangerousCommandGuard classifies command text against configured dangerous-command rules and can present a confirmation prompt. The interactive path is tied to command submission through Chau7, including supported direct terminal input handling.
The guard is not installed inside the shell as a mandatory execution policy. It cannot be described as blocking every command a process, script, remote shell, shell function, or alternate input path may execute.
MCP command submission has a separate filtering and approval flow. Do not claim protection against hidden Unicode, arbitrary multiline scripts, or every paste and wrapper form unless that exact path is verified.
Why it matters
A review prompt can help catch some risky commands, but command-line protection depends on where the command enters and how the shell interprets it. This input guard is not a security boundary.
Frequently asked questions
Does the guard make Chau7 a shell sandbox?
No. It can prompt for supported command submissions, but scripts, child processes, remote commands, and alternate execution paths are outside a universal guarantee.
Does it inspect every pasted or hidden-character command?
The audit does not support complete hidden-character or paste interception claims. Describe only verified input paths.