Block dangerous shell commands and credential file reads before execution.
npx skills add https://github.com/frmoretto/hardstop --skill hardstop
Please help me install this Agent Skill. Skill: hardstop Source: https://github.com/frmoretto/hardstop/tree/main/.claude/skills/hardstop Command: npx skills add https://github.com/frmoretto/hardstop --skill hardstop
This Skill prevents AI-driven or user-initiated shell commands and file reads from causing harm by enforcing a pre-execution safety protocol and guarding against credential exposure.
Instruct the skill to review commands before execution, for example: "Please evaluate the following shell command: rm -rf ~/" and respond with whether to block or allow.
High-intent search queries and answers about installing and using this skill.
Pre-execution safety protocols block dangerous shell commands by applying instant risk assessments and requiring explicit user confirmation before execution, covering Unix-like and Windows environments.
Credential read protection blocks AI from reading sensitive files such as ~/.ssh, ~/.aws, and ~/.env. It warns on risky file reads and prevents access to credential sources to guard against accidental exposure.
Yes, the shell command blocking protocol applies across both Unix-like and Windows environments. It covers commands, file reads, and credential access with layered risk assessment regardless of the operating system.
A pre-execution protocol for AI safety works by performing instant block checks, evaluating risk levels, signaling confidence, and prompting for explicit user confirmation before allowing any potentially dangerous shell command to run.
Yes, LLM-aware safety guidance provides protection when no deterministic hooks are available in the environment. It evaluates commands and shares transparent rationale on why a command is blocked or allowed.
patrykk21Execute literal shell commands directly within the AI interface.
RedHatProductSecurityImplements structured JSON logging and audit trails for MCP servers.
furedeaManage herdr terminal multiplexer workspaces, tabs, and panes via CLI commands.
soy-rafaBlock malicious tool calls and scan installed skills for vulnerabilities.
