Configuring Security Delivery Agents: On Queue, policies and shifts
Reviewed by ContraForce
ContraForce · Help article · Beginner · 5 min read
Set which severities a Security Delivery Agent picks up On Queue, gate its Gamebooks with a confidence threshold, add investigation policies and shifts, link SOPs and check Run History.
A Security Delivery Agent can investigate when someone asks, or pick incidents up on its own. This walkthrough follows the docs' three phases, manual runs, automatic runs by severity, then automatic Gamebooks, and then the policies, shifts and SOPs around them.
What do you need first?
- Agent Center deployed, with at least one agent. Select Agents in the left navigation and follow the prompts to provision the Microsoft Foundry infrastructure. This needs Owner on the Azure subscription.
- To configure the agent: Org Admin or Agent Admin, plus the Owner role on the workspace.
- To manage investigation policies and shifts: Workspace Owner.
How do you run the agent on demand?
- 1. Open an incident.
- 2. Select Actions from the incident toolbar.
- 3. Choose Run agent investigation.
- 4. Review the agent's findings and recommendations.
Run agent investigation is analysis only. Run agent investigation and response also compiles and runs a Gamebook, subject to the workspace's Gamebook settings, and appears only when Allow agent to run gamebooks is on for that severity.
How do you turn on On Queue by severity?
On Queue is fully autonomous: the agent picks up incidents of that severity as they arrive. Manual runs it only when someone triggers it from the Actions menu.
- 1. Select Agents in the left navigation to open the Agent Center.
- 2. Configure the agent from the Agents tab.
- 3. On each severity card (High, Medium, Low, Info), set Mode to On Queue or Manual.
- 4. Set Status after run to No Change, Active, New or Closed.
- 5. Turn on Assign agent to incident if the agent should become the incident's owner.
- 6. Select Save configuration.
New agents start On Queue
New agents default to On Queue. If you aren't ready to automate a severity, set its card to Manual and save.
How do Gamebooks and the confidence threshold work?
Before it acts, the agent builds an investigation graph of the users, devices, IPs, files and emails the incident impacted, then compiles a Gamebook that maps response actions to each entity.
Confidence to respond is set per severity, from 0 to 100. An On Queue run builds a Gamebook only when the agent's malicious confidence meets it; otherwise the response is left to an analyst. Higher is more conservative, and each investigation comment shows the agent's malicious confidence.
- 1. On each severity card, turn on Allow agent to run gamebooks.
- 2. Drag the Confidence to respond slider to your threshold. It must be above 0, or Save configuration stays disabled.
- 3. Select Save configuration.
- 4. In the Workspace Center, open the workspace's Gamebooks tab and turn on Allow service providers to run gamebooks. This needs the Workspace Owner role.
Approval is the default
Allow service providers to run gamebooks is off by default, and while it is off every agent Gamebook waits for approval. Run policies can still block an action or require approval even when the agent is confident.
Turning that setting on also lets the service providers that manage the workspace run Gamebooks there.
How do you escalate True Positives?
By default every investigation ends in the same flow, whatever the verdict. Turn on Advanced options in the agent's configuration and Classification policy cards appear. Each classification gets its own Status after run, Assign agent to incident and optional Custom action, a webhook fired on that verdict.
The pattern the docs suggest: for True Positive, keep the incident open and point a webhook at your on-call or ticketing tooling so your team is paged. For False Positive and Benign Positive, close the incident with no webhook.
How do investigation policies limit the agent?
Investigation policies decide whether an agent investigates an incident at all. Workspace Owners manage them on the workspace's Agents tab in the Workspace Center, and the same list appears on each agent's Policies tab.
Policies allow or block, with no approval option, based on who the incident involves, its title, severity and source. The list is checked top to bottom and the first fit decides. If none fits, the investigation is allowed, so put a narrow Allow above a broad Block for an exception.
A blocked automatic investigation leaves the incident in your team's queue, with the verdict in the workspace audit trail. A blocked manual run shows the person a message naming the policy.
How do agent shifts work?
A shift sets when the agent investigates on its own. It bounds automatic pickup but does not enable it: the severity still needs On Queue.
- 1. Open the workspace's Agents tab, or the agent's Policies tab in Agent Center.
- 2. Click Add policy, then choose Add a shift. The grid opens with Monday to Friday, 9:00 to 17:00 painted.
- 3. Click or drag to paint the working hours.
- 4. Pick the timezone. Hours stay fixed through daylight saving changes.
- 5. Name the shift and save.
Outside the shift, automatic investigations are skipped and are not re-run later, so off-shift incidents belong to your team. Manual runs still work. An Allow above the shift can let specific incidents through, such as titles containing "impossible travel".
How do you link SOPs?
Classification SOPs guide how the agent categorizes severity and priority. Response SOPs guide the containment, investigation and remediation steps it recommends or executes.
- 1. Select SOPs in the left navigation.
- 2. Select Add SOP.
- 3. Choose Classification or Response.
- 4. Upload a Markdown (.md) or plain text (.txt) file and enter a title.
- 5. Click Save.
- 6. Open the SOP, select the Linked Agents tab, click Add Agent and choose the agent.
Links take effect immediately, with no redeploy. SOPs are scoped to a workspace, so upload a shared procedure to each workspace that needs it.
How do you check what the agent did?
In Agent Center, click the Agents tab, the Configure (gear) icon on the agent's row, then Run History. It shows the last 7 days by default: who triggered each run, the incident, severity, tokens and status. For a Failed run, click View error.
An incident that never got an automatic run won't appear there. Check the Mode for its severity, then the workspace's policies and shifts.
What is a sensible rollout?
- 1. Pick one workspace and link the SOPs your team uses most.
- 2. Run the agent manually on a few incidents and compare its verdicts with your analysts'.
- 3. Set Low and Info to On Queue, and leave High and Medium on Manual.
- 4. Add a shift that matches your team's working hours.
- 5. Review Run History and the investigation comments daily.
- 6. Widen in steps: more severities, longer shifts, then more workspaces.
- 7. Turn on Gamebooks last, with a high Confidence to respond and approval still required.
For every setting and the full troubleshooting table, read Configuring Security Delivery Agents in the ContraForce docs.
Keep learning
- Writing incident response SOPs that analysts and AI agents can both follow. A structure for incident response SOPs that analysts and AI agents can follow the same way: decisions as explicit conditions, per-customer exceptions, versioning and a suspicious sign-in example.
- Getting started: connect your first customer workspace. Pre-onboard a customer in the Workspace Center, send the invite, have their admin grant the core and module consents, then confirm the workspace is live with the onboarding checklist.
- How ContraForce automates detection rule tuning for Microsoft Defender and Sentinel. A short explainer on keeping one set of baseline detection rules across every customer workspace, catching drift, and having a Security Delivery Agent review a noisy rule before a person approves the change.
Sources
- ContraForce Docs, Configuring Security Delivery Agents, verified 2026-11-20
- ContraForce Docs, Deploying Agent Center with Microsoft Foundry, verified 2026-11-20
- ContraForce Docs, Workspace Policies for Gamebooks and Agents, verified 2026-11-20
- ContraForce Docs, Agent Shifts: Schedule AI Agent Working Hours, verified 2026-11-20
- ContraForce Docs, SOPs for Security Delivery Agents, verified 2026-11-20
- ContraForce Docs, Agent Run History, verified 2026-11-20