Enforce safety boundaries for destructive commands and directory-scoped edits.
npx skills add https://github.com/2300031147/clawopencode --skill guard-2300031147
Please help me install this Agent Skill. Skill: guard Source: https://github.com/2300031147/clawopencode/tree/main/skills/guard Command: npx skills add https://github.com/2300031147/clawopencode --skill guard-2300031147
Combines destructive-command warnings with directory-scoped edit restrictions to prevent risky changes in production or during debugging, enabling safer operations and reduced blast radius.
Provide the target directory path to activate guard mode and enforce the edit boundary.
High-intent search queries and answers about installing and using this skill.
To enforce edit boundaries, you activate guard mode by providing a target directory path, which restricts file edits to that scope and prevents accidental modifications elsewhere in the file system.
You get warnings for destructive shell commands through PreToolUse hooks that detect actions like rm -rf or DROP TABLE, surfacing clear status messages and requiring explicit confirmation before execution proceeds.
The best way to restrict edits during live production debugging is to activate a full safety mode that combines destructive-command warnings with directory-scoped edit restrictions to reduce blast radius.
Yes, you can integrate safety checks from sibling skills like /careful and /freeze; the system invokes these checks via PreToolUse hooks to enforce safety boundaries across shell and live environments.
Yes, you need to provide a target directory path to activate guard mode, which enforces the edit boundary and ensures destructive commands are intercepted before impacting areas outside that scope.
geoffreywooWarn on destructive commands and restrict edits to a specified directory.
Svenja-devEnforce backups and confirmations before destructive Git operations.
256javyEmit warnings before destructive commands and enforce directory edit boundaries.
SpillwaveSolutionsConfigure hooks.yaml rules to intercept Claude Code events deterministically.
