[Spike] Split-up permission - execute background and foreground

Problem

We have new requirements that expand the scope of role-based permissions for DAP:

Expanded Access Control Requirements:

• Customers want complete control over user access to AI via DAP, including:
  • Agentic Chat in non-namespace contexts (e.g., https://gitlab.com/dashboard/todos)
  • Foundational agents (Planner, Security Analyzer)

Execution Model Complexity:

• We've identified two distinct execution patterns for agents and flows:
  • Background (remote) execution: Requires CI/CD pipelines, consumes compute minutes, currently restricted to Developer+ roles
  • Foreground (local) execution: Runs synchronously without pipelines, currently accessible to Guest+ roles
• Both agents and flows can execute in either mode, but the permission requirements differ based on execution type, not the artifact type (agent vs. flow)

Execution permission split:

We want to split the existing Execute permission into two permissions. Naming TBD.

• Execute foreground for Guest+
• Execute background for Developer+ because of pipeline limitation

Current State:

• Issue #578370 (closed) already mapped locations, permissions, and policies requiring dynamic role-based permissions for background execution.
• The scope now needs to expand to cover foreground execution, Agentic Chat, and foundational agents that are accessible to guest+ roles.

Foreground (IDE) execution set up

• Environment Setup 1: IDE Flows
• #575901 (comment 2913945727)

Proposal

Research and document the current permission controls and code locations for features that support Guest+ access:

1. Agentic Chat - particularly in non-namespace contexts
2. Foundational agents (Planner, Security Analyzer)
3. Foreground (local) execution of custom agents and flows
4. Other DAP features accessible outside namespace contexts or without pipeline requirements.

This research will determine whether we can include the split of the "execute" permission into two separate permissions (execute_background and execute_foreground) in our project or implement a different control mechanism.

Next Steps

• Research existing usage of Agentic Chat with roles lower than Developer (Guest, Reporter)
• Research existing usage of foundational agents with roles lower than Developer
• Research existing foreground (local) execution patterns for agents and flows
• Document code locations that control execution for the above features
• Document current permission checks and authorization logic for Guest+ accessible features
• Determine if execute permission should be split into background/foreground variants
• Assess impact on existing work in #578370 (closed) and parent epic &19601