Knox Developer Safety Pack
The Knox Developer Safety Pack is the product packaging around the Knox Claude Code plugin, CLI, hooks, policy presets, and audit reports. It should help teams answer a simple question:Can we let AI coding agents do real work without giving them unlimited trust?
What It Includes
The pack should include:- Knox Claude Code plugin
- Knox CLI
- default policy presets
- prevention hooks
- audit logs
- reports
- policy export
- MCP tool inspection
- scheduled task inspection
- self-protection against disabling hooks
Core Story
AI developer agents are becoming more autonomous. They can run commands, edit files, install packages, call MCP tools, and schedule work. Knox gives teams a safety layer around those actions:- check before execution
- block dangerous operations
- sanitize risky commands where possible
- audit allowed and denied actions
- detect repeated probing
- protect policy files and hooks
Recommended Demo Flow
Use this sequence for demos, screenshots, or internal testing.1. Show Status
2. Test Safe Command
3. Test Risky Command
4. Show Audit
5. Show Report
6. Export Policy
MCP/CLI-First Integration Strategy
Knox fits the Qoris MCP/CLI-first direction. Instead of building every external integration as custom backend code, agents can use MCPs and CLIs. Knox then governs the action path:- CLI command safety
- MCP input inspection
- file and config protection
- audit events
- escalation after repeated denials
Example Policy Outcomes
What Eliel Can Work On
Eliel can work on product-facing assets without changing the core policy engine:- Developer Safety Pack copy
- docs examples
- screenshots from
knox status,knox audit, andknox report - MCP/CLI-first safety messaging
- marketplace/plugin positioning
- demo script
- comparison against “agent without Knox”
