Intercept destructive commands and restrict file writes to specified directories.
npx skills add https://github.com/samymity/node-poc --skill safety-guard-samymity
Please help me install this Agent Skill. Skill: safety-guard Source: https://github.com/samymity/node-poc/tree/main/.claude/skills/safety-guard Command: npx skills add https://github.com/samymity/node-poc --skill safety-guard-samymity
When working on production systems, running AI agents in fully autonomous mode, or performing sensitive operations like database migrations and code deployments, accidental execution of destructive commands (such as rm -rf, git push --force, or DROP TABLE) or unintended file edits can lead to catastrophic data loss, extended system downtime, and security vulnerabilities. This Skill eliminates these risks by providing lightweight, configurable protection layers that integrate directly with agent tool execution.
Enable the Safety Guard skill to automatically block high-risk destructive commands and restrict file write operations to your specified project directory for safe autonomous AI agent workflows.
High-intent search queries and answers about installing and using this skill.
To prevent autonomous agents from running destructive commands like rm -rf or git push --force, you can enable Careful Mode to intercept high-risk commands, explain their impact, and require explicit user confirmation before execution.
You can restrict file edits to a specific directory by enabling Freeze Mode, which locks all file write and edit operations to a designated directory tree and blocks any modifications outside the allowed path.
The best way to protect production systems during autonomous code deployments and database migrations is to use Guard Mode, which combines command interception with directory-scoped write restrictions to block destructive operations across the entire system.
Yes, you can block destructive system commands while allowing unrestricted read access to the entire codebase by using Guard Mode, which restricts file writes to a defined directory and intercepts dangerous commands globally.
Yes, the safety guard provides safer alternatives when it intercepts a dangerous command, offering suggestions for common destructive patterns like rm -rf, git reset --hard, docker system prune, and npm publish during the confirmation prompt.
Use Careful Mode when you need to intercept and confirm destructive commands without restricting file write locations, and use Freeze Mode when you specifically need to lock file modifications to a designated directory tree for focused autonomous operations.
smarmen86Intercept destructive shell commands and restrict file writes to a defined directory.
jonnabioGuide reversible data and schema migrations with rollback planning.
Geralt-bytePerforms GitHub PR checks, workflow runs, and API queries via gh CLI.
hypeJunctionBlock destructive system and database commands via a Bash tool use hook.
