Collection
Let AI do useful work without giving it accidental authority
Design AI-assisted workflows so tools, permissions, evidence, approvals, testing and monitoring match the real task and limit the blast radius of a bad output.
15 elements · Explore in any order.
4 checklists · 6 principles · 5 protocols
15 elements
Give the agent only the tools this job needs
Every extra tool is another verb the system can accidentally conjugate.
Read the idea PrincipleStart an agent read-only when writing is not required
Reading first gives you evidence. Writing first gives you evidence plus cleanup.
Read the idea PrinciplePrefer a task-shaped tool to an open-ended one
A function named `create_invoice_draft` has fewer dangerous interpretations than `run_anything`.
Read the idea PrincipleRun agent actions in the user's authorization context
A helpful assistant should not quietly become a shared administrator.
Read the idea ProtocolPut approval immediately before the high-impact action
Approval at the start of a long plan is not approval of the action the plan eventually invented.
Read the idea PrincipleTreat retrieved content as data, not as new authority
The page you asked the agent to read should not get to rewrite the job description.
Read the idea ChecklistKeep secrets out of the system prompt
A prompt is a poor vault, even when nobody intends to show it.
Read the idea PrincipleEnforce permissions outside the model
A sentence in a prompt is guidance. An authorization check is a control.
Read the idea ProtocolValidate model output for the system that will consume it
Readable text becomes a different risk when the next component treats it as code.
Read the idea ChecklistLog what the agent did, not only what it said
The chat transcript is not the audit trail if the real effects happened elsewhere.
Read the idea ProtocolCross-check the fact before the agent acts on it
Confidence is a writing style. Evidence is a different object.
Read the idea ProtocolKeep an eval set that can embarrass the agent
A demo asks whether the system can succeed once. An eval asks where it reliably fails.
Read the idea PrincipleTest the agent under conditions that resemble the real job
A perfect lab result can still be a poor rehearsal.
Read the idea ProtocolMonitor the agent after deployment
Deployment is the first time the model meets all the mess your test set forgot.
Read the idea ChecklistBound how much damage one agent run can do
Automation turns one mistake into throughput unless you design a brake.
Read the idea