All ArticlesRuntime Truth
Access & Permissions
Feature Blog

AI Agent Posture Management: Who Built What and How It's Configured

Learn how AI agent posture management reveals who built each agent, how it's configured, and what it can actually access, before a single action causes harm.

Obsidian Editorial Team
Security Research
·
Obsidian Security
·
May 26, 2026
May 28, 2026
Key Takeaways
  • One enterprise discovered 377 Microsoft Copilot agents running in their environment that their security team had never seen, never approved, and could not attribute to any owner.
  • Every one of those agents had inherited credentials.
  • Every one could act on behalf of its creator.
  • That is the posture problem, and it starts before any agent takes a single action.
  • AI agent posture management is the practice of answering four questions about every agent in your environment: who built it, how is it configured, what can it actually access, and does any of that align with your security policy?
  • In 2026, most security teams cannot answer them for more than a fraction of their deployed agents.

Why AI Agent Posture Management Exists

Security teams did not create an agent inventory problem. Business teams did, at speed, with tools that made it easy. Platforms like Microsoft Copilot Studio, Salesforce Agentforce, and n8n allow non-technical users to build and deploy agents in hours. No change ticket. No security review. No record in any system the security team controls.

The result is shadow AI at the agent layer. Not just shadow AI apps that receive data passively, but shadow agents that take autonomous actions, hold embedded credentials, and connect to production SaaS systems. An agent is not a browser tab. It is a machine identity with a blast radius.

Traditional security tools were not designed for this. Network and endpoint solutions cannot see agent activity across SaaS platforms. Identity providers track human logins, not agent invocations. Native platform logs exist but are siloed per tenant, require manual correlation, and do not answer the question security teams actually need answered: what can this agent actually do?

That gap is where AI agent posture management begins. Before any runtime enforcement conversation is possible, security teams need a complete, accurate picture of what agents exist, who owns them, and how they are configured. You cannot govern what you cannot see.

The Four Posture Questions Every Security Team Must Answer

Posture management for AI agents resolves around four core questions. Each one builds on the last.

Who Built This Agent?

Creator attribution is the foundation of agent governance. Every agent has an owner: the person or service account that created it, whose credentials it may be using, and who is responsible for its behavior. Without creator attribution, there is no accountability chain.

Creator attribution also surfaces orphaned agents. When a creator's account is disabled (through offboarding, role change, or account cleanup), the agent does not stop. It continues running with the creator's inherited credentials. The creator's access level, scope, and permissions remain active inside the agent. Security teams that track creator accounts can identify these orphaned agents immediately. Teams that do not track them are running agents with credentials attached to identities that no longer exist in their IAM system.

How Is This Agent Configured?

Configuration covers the agent's connection type, authentication method, accessible data sources, invocation scope, and tool connections. Each configuration choice carries a distinct risk profile.

Key configuration signals include:

  • Authentication mode: is the agent using maker mode (fixed creator credentials) or user context (invoker's own permissions)?
  • Invocation scope: is the agent accessible only to specific users, to the entire organization, or publicly via a URL?
  • Connected data sources: which SaaS applications, knowledge bases, and MCP servers does the agent connect to?
  • Credential handling: are credentials embedded in the agent configuration, or managed through a secrets system?

What Is This Agent's Effective Authority?

Theoretical configuration is what the agent is set up to do on paper. Effective authority is what the agent can actually do inside each connected SaaS application after all entitlements resolve.

These two things are frequently not the same. An agent may appear to have limited configuration while holding OAuth grants to a CRM system where the creator holds admin-level access. The configuration looks fine. The effective authority is critical-severity.

Mapping effective authority requires correlating agent configuration with actual entitlements from connected applications. This is the step that most agent inventory tools skip, and it is the step that matters most for understanding real blast radius.

Does Any of This Align With Policy?

Once the first three questions are answered, posture management applies policy logic. Is this agent using an authentication mode your organization has approved? Does its invocation scope match its intended use case? Are its connected data sources within the scope of what the agent's business purpose requires?

Policy alignment checks produce the risk signals that drive remediation. They also produce the audit evidence that security teams need for compliance reporting and board-level conversations.

High-Risk Configuration Patterns: What Posture Reveals

Across enterprise AI agent deployments, several configuration patterns consistently produce elevated risk. Understanding the mechanism behind each one is essential for prioritizing remediation.

Maker Mode with Sensitive Data Access

Maker mode is the default configuration in many agent builder platforms. When an agent is built in maker mode, it uses the creator's credentials for every action it takes, regardless of who invokes it. A user with no Salesforce access can invoke a maker-mode agent built by a Salesforce administrator. That user now has effective access to CRM data at the administrator's privilege level. No IAM rule was violated. The agent performed exactly as designed. Your access controls were bypassed.

This is the machine insider risk pattern. The agent acts like an insider with elevated credentials, but no insider risk program covers it.

  • What the control enforces: maker mode detection flags every agent where the creator's credential level exceeds the invoker's provisioned access, surfacing the privilege gap before it is exploited.
  • Why theoretical configuration fails: no configuration file reveals that a "read" connector resolves to admin-level authority inside the connected SaaS application after OAuth grant resolution.

Publicly Accessible Agents

An agent configured for public access can be invoked by anyone with its URL, including unauthenticated external users. This configuration is sometimes intentional for customer-facing use cases. It is frequently unintentional for internal workflow agents that were published without scoping the invocation audience. Posture management surfaces every publicly accessible agent and flags the combination of public access with sensitive data connections as a critical-priority finding.

  • What the control enforces: invocation scope review ensures every public agent has a documented business justification and limited data access.
  • Why theoretical configuration fails: public agents are often created as internal prototypes and never re-scoped, meaning the publicly accessible configuration persists long after the original use case changes.

Org-Wide Accessible Agents with Broad Permissions

An agent accessible to every user in the organization is not inherently dangerous. An agent accessible to every user in the organization while holding admin-level credentials to a financial system is a different matter entirely. Posture management identifies this combination and scores it accordingly.

  • What the control enforces: combined scoring of access scope and permission level, not individual factors in isolation.
  • Why theoretical configuration fails: org-wide accessibility and admin-level credentials each score as medium severity independently, masking the critical blast radius of their combination.

Hardcoded Credentials

Some agent configurations embed credentials directly in workflow definitions or connection strings. These credentials do not rotate. They do not expire. If the agent configuration is exported, shared, or accessed by an unauthorized party, those credentials travel with it. Platforms like n8n and others that support workflow-based agent building are particularly susceptible to this pattern.

  • What the control enforces: credential hygiene scanning across all agent configurations, flagging any embedded credential not routed through a secrets management system.
  • Why theoretical configuration fails: hardcoded credentials appear legitimate in configuration reviews because the agent is technically authorized. The risk is in the credential's exposure surface, not its validity.

Agents with Disabled Owners

An agent whose creator account has been disabled continues operating with that creator's permissions. This is the orphaned agent problem. The human identity no longer exists in the IAM system, but the agent's effective authority remains intact. Posture management surfaces every agent where the creator account is disabled, treating these as high-priority findings regardless of the agent's other configuration signals.

  • What the control enforces: real-time correlation between agent creator accounts and identity provider status, flagging every agent where the owner account has been deprovisioned.
  • Why theoretical configuration fails: configuration reviews show the agent as "active" and credentials as "valid." The orphaned status is only visible when creator identity is correlated against a live identity provider.

Toxic Combinations: When Risk Factors Stack

Individual risk factors are important. Combinations of risk factors are critical.

A single medium-severity finding (such as an agent accessible org-wide) may not require immediate escalation. That same agent, when it also uses maker mode with admin credentials and connects to a knowledge base containing sensitive financial records, is a different situation entirely. The risk factors compound. The blast radius expands. The priority becomes critical.

This is the concept of toxic combinations in AI agent risk management. Posture management systems that score individual findings in isolation miss the compounding effect. Effective posture management correlates multiple signals per agent and surfaces the combinations that represent the highest actual risk.

Common toxic combinations:

Risk Factor ARisk Factor BCombined SeverityMaker mode with admin credentialsOrg-wide or public accessibilityCriticalOrphaned agent (disabled owner)Active connections to sensitive dataCriticalShadow agent (unmanaged, no security review)Broad OAuth scopes to production systemsCriticalHardcoded credentialsPublic webhook endpointCriticalConnector to unregistered applicationMaker mode authenticationHigh

Prioritizing remediation by toxic combination score, rather than individual finding severity, focuses security team effort on the agents that represent the greatest actual exposure.

Posture Is the Starting Point, Not the Finish Line

Here is the limitation every security team needs to understand clearly: posture management shows you the theoretical configuration. It does not show you what the agent actually did.

Configuration is not reality.

An agent can be correctly configured according to every policy your team has defined, and still take actions at runtime that produce unauthorized data access. Probabilistic agents do not always behave as their configuration predicts. Action chaining (where an agent sequences multiple tool calls across applications) can compound access in ways that no static configuration review anticipates. A confused deputy scenario (where an agent with elevated permissions is manipulated into performing actions for an unauthorized user) does not show up in posture data.

This is why posture management is the required first step, not the complete answer. Security teams that stop at posture are still ghost chasing: reviewing theoretical configuration signals with no evidence of what actually happened at runtime. The question posture answers is "what could this agent do?" The question runtime visibility answers is "what did this agent actually do, who triggered it, and what data did it touch?"

Both questions matter. Posture management without runtime visibility leaves a gap. Runtime visibility without posture context lacks the attribution layer needed to understand why an agent had the access it used.

For a deeper look at how runtime visibility extends beyond static posture, see the AI agent security best practices guide.

Building a Single Pane of Glass for Agent Posture

The operational challenge of AI agent posture management is not conceptual. It is mechanical. Security teams managing agents across Microsoft Copilot Studio, Salesforce Agentforce, Amazon Bedrock, n8n, and ChatGPT Enterprise are dealing with siloed inventories, inconsistent risk signals, and no unified view of cross-platform agent authority.

A single pane of glass for agent posture requires:

  • Cross-platform agent discovery: every agent across every supported platform, surfaced in one inventory regardless of which business team built it.
  • Creator and owner attribution: each agent linked to its creator account, with flags for disabled or unrecognized owners.
  • Configuration risk scoring: each agent scored against policy, with toxic combination logic applied across multiple risk factors simultaneously.
  • Effective authority mapping: each agent's actual access scope inside connected SaaS applications, not just its theoretical configuration.
  • MCP server inventory: every MCP server connection each agent holds, including connections to unregistered or shadow applications.

This is the foundation that makes every downstream security conversation possible. Governance, access control, and least privilege enforcement for AI agents all depend on having an accurate, complete, and continuously updated picture of what agents exist and what they can actually do.

For teams managing Microsoft Copilot or Salesforce Agentforce deployments specifically, platform-specific posture risks compound quickly at scale. Starting with a structured AI agent risk assessment gives security teams a baseline inventory and a prioritized remediation list before the agent count grows further.

Frequently Asked Questions

No items found.