iiRecord
Agentic AI Atlas · GAP-SEC-003: Permission Request and Denial Hooks
page:docs-harness-features-backlog-gaps-security-gap-sec-003a5c.ai
II.
Page reference

page:docs-harness-features-backlog-gaps-security-gap-sec-003

Reading · 1 min

GAP-SEC-003: Permission Request and Denial Hooks reference

Typed permission hooks that distinguish between different kinds of human interaction: clarification requests, security approvals, intervention alerts, and notifications. Different interaction types get different UX flows.

Pagewiki/docs/harness-features-backlog/gaps/security/GAP-SEC-003.mdOutgoing · 0Incoming · 1

GAP-SEC-003: Permission Request and Denial Hooks

FieldValue
Categorysecurity
PriorityHigh
EffortL
StatusPartial

Description

Typed permission hooks that distinguish between different kinds of human interaction: clarification requests, security approvals, intervention alerts, and notifications. Different interaction types get different UX flows.

Current State

All human interaction flows through the same breakpoint mechanism. A breakpoint for "confirm destructive action" looks the same as one for "please clarify requirements." The interaction module provides input widgets but not semantic interaction types.

Target State

InteractionKind enum: clarification, approval, intervention, notification, handoff. Each kind routes through different UX flows with appropriate urgency and context. Permission denied events logged for audit.

Dependencies

  • GAP-SEC-001 -- governance policy for permission evaluation

Key Files

ComponentPath
Breakpoint typespackages/sdk/src/breakpoints/types.ts
Interaction modulepackages/sdk/src/interaction/

Recommendation

Phase 2 implementation. Introduce InteractionKind to the breakpoint system. Route different kinds through different UX flows. Backward compatible with existing breakpoints.