Acerca de
HardStop is a comprehensive safety layer for Claude Code and agentic AI environments, designed to prevent catastrophic commands and sensitive data exfiltration. It implements a rigorous pre-execution protocol that categorizes actions into safe, risky, or dangerous tiers, ensuring that operations like recursive deletions, system configuration changes, and credential file access are caught and blocked or require explicit user confirmation. By providing LLM-level awareness of command intent, HardStop acts as a mechanical brake for AI-initiated actions, shielding developers from hallucinations, mistakes, and prompt injection attacks.
