Review overall risk posture
Use Overview in Protect to assess risk across your published agents and Glean Assistant, then decide what to review next. It combines an overall risk score, a historical trend, recommended remediations, and a list of your riskiest agents.
Prerequisites
Complete the following prerequisites:
- Confirm that agent posture is enabled for your environment. If Overview doesn't appear in Protect, contact your Glean account team.
- Have the Super Admin or Sensitive Content Moderator role. See Administrator Roles.
- To review every listed agent, have permission to view all agents, such as the Super Admin or Agent Moderator role. Otherwise, the current score and agent list include only the agents you can access.
Open the overview
Open Protect from the Glean navigation, then click Overview. Check the As of timestamp to see when Glean computed the displayed posture.
The agent part of the current score and the agent list reflect the published agents you can view, not every agent in your organization. Agents that have never been published and unlisted agents don't contribute. The Glean Assistant score and the trend cover your whole deployment.
Interpret the overall score
The Overall risk score runs from 0 to 100, and a higher score indicates more risk. It combines two scores:
- The average score of the agents you can view that have a score, including partially assessed agents. Agents without a score don't count toward the average.
- The Glean Assistant score, which is based mostly on its recent open runtime issues. See Review Glean Assistant security posture.
When both scores are available, the agent average makes up 70% of the overall score and the Glean Assistant score makes up 30%. When only one is available, the overall score equals that score. The score shows Includes Glean Assistant when the Glean Assistant score counts.
A dash means that neither score is available yet. It doesn't mean that your agents and Glean Assistant have no risk.
An agent's score combines configuration findings, risks in its tools, skills, and subagents, and open runtime issues. The overall score can therefore change as agents receive scans, as runtime issues open or close, or when the set of agents you can view changes.
Read the risk trend
The trend plots the overall risk score for each day of the past week from saved daily snapshots, not live findings. Hover over a day to see its score and whether it Includes Glean Assistant or covers Agents only. Days without a recent Glean Assistant score count only agents.
The trend counts agents across your deployment, including agents you can't view. If you can view only some agents, the trend can differ from the current score beside it. No trend data yet means that Glean hasn't saved a scored snapshot for the displayed period.
Review the riskiest agents
The Risky agents section shows up to six of the highest-risk assessed agents you can view. Each entry includes its score, risk summary, and creator.
Click View all agents to open the inventory. Use the inventory filters to find an agent, then click its name to review its configuration findings, runtime issues, and component risks. See Review agent security posture.
Choose a remediation
Expand an item in Top recommended remediations to read the affected guardrails, creators, or scan rules. Glean shows a recommendation only when its conditions are met, so the list can vary between environments and users.
The available recommendations are:
- Enable guardrails across all agents identifies built-in guardrails that don't cover every agent in your view. Click View policies to review their configuration.
- Review agents owned by high-risk users identifies creators who own at least two agents above the risk score threshold shown in the recommendation. The recommendation concerns the agents' risk, not a separate risk assessment of the creators. Click View agents to open the inventory.
- Review rules causing agent scan failures identifies scan rules with high or critical findings across at least two agents. When available, the recommendation estimates the average score reduction from resolving those findings. Click View agents to investigate.
The recommendations link to the relevant pages but don't change guardrail settings or agent definitions. View agents opens the full inventory, so use its filters to find the affected agents.
No recommended remediations right now means that no recommendation meets the current conditions. Keep reviewing individual agents, because their findings can still show risks that need attention.
See also
- AI Security Insights — review guardrail activity and runtime security trends
- Configuring Policies — change the guardrails that apply to AI interactions