Sub-agent permissions

A sub-agent grant is a narrower subset of its parent’s permitted operations, resources, duration, purpose and delegation depth.

01 / The operating case

What changes in a real workflow?

A research child may retrieve three approved files and summarize them; it cannot send messages or pass file access to another child.

Trace the authority boundary
  1. 01Origin grant
  2. 02Parent agent
  3. 03Narrow child grant
  4. 04Action boundary

02 / Decision contract

What the executor must check

At handoff, compare child scope to parent scope and make resource, verb and further-delegation rights explicit.

InputPrincipal · task · action · target
DecisionGrant ancestry · policy · current state
ResultEnforce · record · verify outcome

Architecture review / negative test

Make the boundary testable.

A child needs only the operations, target set and duration required for its assigned subtask. Review fan-out carefully: multiple child grants should not each inherit the full parent budget or a right to delegate onward. A subset validator should compare structured semantics, not text labels that look narrower.

03 / Failure and evidence

What goes wrong, and what can be proven?

Failure mode

A copied parent token grants the child unrelated send or delete capabilities.

Evidence to retain

Store grant edge, subset check, issuer, child identity and expiry.

04 / Canonical scope

Why this reference stands alone

Multi-agent authorization owns the final action decision; this page owns grant construction at handoff.

Primary references

Read the underlying material

Architecture discussion

Choose one consequential action and make its boundary explicit.

Request a Conversation