tp9imka gates held 0 CV
<= all systems
Agent securitysecurity

Go on. Read my inbox.

Policy in code, not in the prompt. An allowlist hook on every tool call, message text treated as data, a merge guard that fails closed, and runtime protection that detects the act of hiding.

inbox agent · policy hookSIMULATION
guard: denied rule="message text is data" tool=policy_override => not executed, logged

Constraint

An inbox agent reads untrusted messages and can send. A prompt that says “be careful” is not a control.

What it refuses

  • Any tool not on the allowlist. A hook runs before every tool call, and every decision is logged.
  • Instructions inside a message. Message text is data.
  • Sending outside a 1-1 chat. Checked in two places in code.
  • A merge past an unreadable check. The guard fails closed.

Also

An AI assistant inside encrypted groups joins as a visible member with its own key and read-only tools. Removing it revokes access cryptographically. On mobile, runtime protection detects the act of hiding, not the name of the hider.

What broke

Story pending.

The wall of no

Every shortcut has a gate.

These are real rules from the crews I run, not a demo. A no needs a gate. A yes needs an undo.

GateTriggerAction
REVIEW VERDICT NEW COMMIT PUSHED VOID, JUDGE THE NEW HEAD
QA VERDICT BOUND TO AN OLD SHA VOID, NO RECEIPT
MERGE GUARD CHECK STATE UNREADABLE FAIL CLOSED
SAME CHECK RED 3 HEADS IN A ROW DRIVER PARKS, RE-PLANS
RED CI NEW HEAD WAKE IMPLEMENTER ONCE
INBOX AGENT TOOL NOT ON ALLOWLIST DENIED, LOGGED
INBOX AGENT SEND OUTSIDE A 1-1 CHAT REFUSED, CHECKED TWICE
BUG REPORT SAME FINGERPRINT MERGED INTO ONE ISSUE
REQUIREMENT CRITERIA HASH DRIFTED RE-SCORED OVERNIGHT
PRODUCT QUESTION INSIDE SCOPE DECIDED FOR IVAN - UNDO
SCOPE CHANGE NEW COMMITMENT WAKES IVAN
CREW COST PER TICKET NOT MEASURED YET