KIT 001 public-safe
AI Agent Safety Checklist
A first-pass operator checklist before connecting an agent to tools, inboxes, files, websites, money or customers.
#agents #safety #tools
Public-safe kits · no private wiring diagram
Download the working discipline.
These are not secret playbooks. They are public-safe checklists and templates: enough to help a serious operator build safer systems, without exposing private infrastructure, customers or internal routes.
KIT 002 public-safe
Human Approval Gates Template
A reusable gate map for deciding what an agent may draft, propose, execute or never touch.
#approval #workflow #risk
KIT 003 public-safe
Local AI Kill-Switch Checklist
A practical checklist for putting the stop path outside the model/provider/action loop.
#local-first #control #fallback
No credentials, internal hostnames, customer examples, exact private deployment routes, paid-data workflows, or live-control instructions. The useful bit is the judgement: what to gate, what to log, what to keep local, and where to stop before clever turns expensive.