agent-identity-privilege-reviewer

Review agent identity architectures for privilege escalation and confused-deputy risks.

2|Updated Jul 6, 2026
One-click install
npx skills add https://github.com/nguyenpv1980-wq/Project-Aegis --skill agent-identity-privilege-reviewer
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-identity-privilege-reviewer
Source: https://github.com/nguyenpv1980-wq/Project-Aegis/tree/main/.claude/skills/agent-identity-privilege-reviewer
Command: npx skills add https://github.com/nguyenpv1980-wq/Project-Aegis --skill agent-identity-privilege-reviewer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill reviews how agents and sub-agents are identified, what privileges they hold, and how authority moves through delegation chains so you can catch escalation paths before they become incidents.

Core Features & Use Cases

  • Identity inventory: Map each agent to the identity it runs under and flag shared, ambient, or over-broad credentials.
  • Privilege analysis: Compare what each identity can do against what the agent actually needs, including scopes, roles, and tenant reach.
  • Delegation and confused-deputy review: Trace user-to-agent-to-sub-agent-to-tool authority flow, verify attenuation, and detect privilege laundering or cross-tenant abuse.
  • Dual attribution checks: Confirm that side effects can be attributed to both the human principal and the acting agent, not just a service account.
  • Use Case: A platform team asks whether its orchestrator can safely spawn workers with narrower, task-scoped credentials instead of reusing one powerful token across the whole fleet.

Quick Start

Review the agent identity architecture for shared accounts, privilege excess, delegation attenuation failures, confused-deputy risks, and missing principal-plus-agent attribution.

Frequently Asked Questions about agent-identity-privilege-reviewer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I identify confused-deputy risks in multi-agent delegation chains?

Identify confused-deputy risks by tracing user-to-agent-to-sub-agent authority flow, verifying privilege attenuation at each delegation hop, and detecting cross-tenant privilege laundering or escalation paths. Compare actual agent scopes against least-privilege needs to flag excess authority.

What is the best way to audit privilege escalation and agent identity in multi-agent systems?

Audit privilege escalation by mapping each agent to its underlying identity, comparing held scopes against required task scopes, and reviewing dual attribution logs to ensure side effects trace to both the human principal and the acting agent, not just a service account.

When do I need to review least-privilege credentials for task-scoped agent workers?

Review least-privilege credentials when an orchestrator spawns sub-agents or workers, especially if reusing standing tokens across a fleet. Compare task-scoped credential needs against actual agent scopes to prevent privilege drift and cross-tenant abuse.

How do I check if my agents are using shared or ambient credentials instead of task-scoped tokens?

Check for shared or ambient credentials by performing an identity inventory that maps each agent to its running identity, then flagging over-broad, shared, or ambient credentials that exceed the specific task-scoped least-privilege requirements.

Does confused-deputy review work for cross-tenant delegation and authority flow analysis?

Confused-deputy review works for cross-tenant delegation by tracing authority flow through delegation hops, verifying attenuation, and detecting privilege laundering. It confirms side effects are attributed to both human and agent principals across tenant boundaries.

Why does my multi-agent audit logging only show service accounts instead of the human principal?

Audit logging only shows service accounts when dual attribution checks are missing. Review agent identity architecture to confirm side effects are attributed to both the human principal and the acting agent, preventing obscured accountability in delegation chains.