Reins adds deterministic PreToolUse and PostToolUse security hooks for agent runtimes. It enforces rules (allow/warn/block) for shell commands, file edits, and MCP calls, logs audit entries, and can integrate with Watchtower for centralized policies and immutable admin rules.
Use Reins whenever an agent has permission to run shell or file operations, access external MCPs, or perform potentially destructive work — especially in production, shared, or sensitive environments. It's also useful during audits or when enforcing org-wide safety policies.
Best suited for Claude Code and other agent runtimes that support pre/post tool hooks and shell interception.
Reins is a runtime security hook system for agents that enforces policies on shell commands, file operations, and MCP calls. The SKILL.md is thorough with clear CLI reference, hook behavior documentation, and response format templates. However, the core npm package @pegasi/reins does not exist on the registry yet, making the skill completely non-functional. The install script failed with a 404. Concept is strong and timely but currently vaporware.
@pegasi/reins (npm package not yet published)Well-documented skill with a genuine security governance concept. The auto-update mechanism (reins upgrade) and Watchtower phone-home are mild security concerns but aligned with the product's purpose. Main issue is the package doesn't exist yet — the skill is aspirational documentation rather than usable software. Would benefit from references/ directory for policy examples and separating the CLI reference into its own file.