Your team has written the operating procedures. Built the workflows. Defined how agents should approach the work. But every new agent still means copying instructions, and people who need automation still need help finding and running it.
Itential Platform 6.7 addresses both sides of that problem. Agent Skills give FlowAgents reusable domain knowledge they can load when needed. Service Requests gives people a dedicated place to request automation and track the result.
Together, they make more of what your team knows and builds available to the people who need it. This release also brings more precise agent inputs, simpler device targeting, consistent query syntax, and improvements to integration management.
An agent needs more than access to tools. It needs instructions for how to approach the work: what to check, which procedure to follow, and how to interpret what comes back.
Putting all of that into every agent’s system prompt creates a maintenance problem. The same knowledge gets copied across agents. Updates have to be repeated. Instructions for tasks an agent may never perform still occupy its context.
Agent Skills package that knowledge separately so it can be reused across FlowAgents. Each skill includes a SKILL.md file and can include supporting reference and asset files.
At the start of a session, the agent sees the names and descriptions of its attached skills. When a skill is relevant, the agent loads its full instructions. Supporting files can be loaded as needed. The detailed procedures stay out of the initial context until they are useful.
For example, a team could capture its procedure for investigating a connectivity problem in a skill and attach it to multiple agents. Each agent can load the procedure when that work comes up. The prompt defines the agent’s role and rules; the skill supplies reusable knowledge for the task.
Builders upload skills to a shared library and attach them in FlowAgent Builder. Permissions control who can access and maintain each skill.
An attachment can stay pinned to an immutable version or follow the latest version. For attachments following latest, the version resolves when a session starts. Uploading a new version does not change the instructions in an already running session.
Agent Sessions shows which skill versions loaded and whether each load succeeded, giving builders a record of the instructions available during execution. Scripts included in a skill do not execute in this release.
The Agent Skills documentation also introduces Infrastructure Agent Skills, an open source library of reviewed network and infrastructure skills. Teams can upload these skills into the platform and adapt them to their own procedures.
This library gives teams a starting point for building reusable operational knowledge. These are Agent Skills used inside FlowAI. Itential AI Skills for external AI tools remain a separate offering.
The person requesting a service should be able to find it, provide the required information, and see what happened.
Operations Manager gives administrators the controls and execution detail they need to manage automation. Someone requesting a report or a network change has a different job to do.
Service Requests gives those users a dedicated place to request services backed by workflows or FlowAgents.
Users see the services they are authorized to access, choose one, and complete its form. They can run it immediately or create a recurring schedule, then review request status, submitted inputs, and available output.
Existing manual triggers appear as services for authorized users. Administrators continue managing those triggers and their underlying automations in Operations Manager. API and event triggers do not appear in the service catalog.
Access combines the Service Requests and trigger application permissions with the underlying automation’s read and write groups. Read-group access allows users to request a service and create a schedule. Write-group access is required to edit, pause, resume, or delete schedules. Administrators should review those groups before opening services to additional users.
Requesters cannot edit the underlying workflow, FlowAgent, trigger, or shared defaults through Service Requests. Requests cannot be canceled from this experience.
A person requests automation in Service Requests. If the automation pauses for approval, review, or additional input, an operator handles the task in Work Center. The requester continues tracking status in Service Requests.
Administrators continue using Operations Manager to manage and run automations and investigate workflow execution. Agent Sessions provides the detail needed to investigate FlowAgent execution.
The same person may perform several of those jobs. Each experience gives them the information and actions appropriate to the work in front of them.
Service Requests is included in the Itential Platform at no additional cost. Customers requesting automation through ServiceNow can continue using that path. Service Requests adds a native option for teams that need one.
Builders have more control over what a FlowAgent accepts. In FlowAgent Builder, inputs can use types including strings, numbers, booleans, objects, arrays, and null, with multiple accepted types where needed.
Inputs and object keys can be optional, and descriptions explain what each field expects. Top-level validation supports string lengths and formats, numeric bounds, and integer-only values. Formats include dates, times, email addresses, IPv4, and IPv6.
Object inputs can define keys one level deep. Arrays can specify an item type. Validation rules apply to top-level inputs; nested keys and array items do not have the same rule controls.
That lets a builder describe a request more precisely and reject invalid input before execution begins. Builders should also keep the agent’s prompt aligned with the input schema.
Operators already organize devices with Inventory Manager tags. Platform 6.7 lets them use those tags when creating a Configuration Manager Enterprise device group or adding devices to an existing group.
In the interface, operators can search across inventories, filter by tags, and review matching devices and counts before saving. They can remove individual devices from a selection without starting over. Device detail pages also show Inventory Manager tags.
The resulting group captures membership at save time. Later tag changes do not automatically change the group.
The corresponding Studio tasks also support tag-based selection, with each task targeting one inventory. Workflows spanning multiple inventories need a separate task for each additional inventory. Review the device group guidance, including execution-account access, before automating group updates.
A workflow task returns a large object. The next step needs one value buried inside it. Builders can use a task input query to extract that value where it is consumed, or a Query task when they need to produce a reusable result.
Platform 6.7 gives both approaches the same full query syntax, including filters and helper functions in task input queries. The new :getLiteral() function provides literal-value handling. The Query reference documents the syntax and examples.
Some older property-access forms in task input queries are deprecated. Existing queries continue to run, and affected inputs display a warning with an Update action. Conversion is opt-in and takes effect when the workflow is saved.
Back up workflows before converting affected queries. The conversion cannot be reversed through the update action, and converted syntax may not work on older releases. Review the deprecation notice when planning an upgrade.
Integration Models and Integration Instances move to the Mosaic interface, bringing a consistent layout to configuration and troubleshooting.
Administrators can review connection settings, gateway settings, variables, and JSON configuration in the refreshed instance pages. Logging improvements include syslog support and more reliable application of configured log levels. These logging controls are not available for integration instances that run through a Gateway.
These changes put frequently needed configuration and troubleshooting controls closer to the administrator doing the work.
The accompanying Itential Gateway 5.5.4 release adds support for custom RPM installation directories, fixes secret placeholder resolution in Basic and Proxy Authorization headers, and updates dependencies to address security vulnerabilities.
Platform 6.7 introduces Trigger Service, which takes over trigger execution previously handled by Operations Manager. Avoid creating, editing, or deleting automations and triggers during the upgrade, and allow the migration to finish before managing triggers.
Administrators should also review audit logging. Platform 6.7 honors audit_enabled, which defaults to false. Environments that need to continue writing audit.log must explicitly enable it on each node at startup.
The November 11 changelog includes the full upgrade actions, including Keycloak issuer matching, Redis Sentinel permissions, and subscriber filters affected by Trigger Service identifiers.
Agent Skills make domain knowledge reusable across FlowAgents. Service Requests makes workflows and FlowAgents accessible through a common request experience. The supporting improvements reduce the effort required to define inputs, target devices, query data, and manage integrations.
That gives teams a practical path from capturing expertise to delivering a service someone else can use.
Explore the Platform 6.7 feature announcement, and talk to your Itential team about putting Agent Skills and Service Requests to work in your environment.
See how Itential connects AI reasoning to governed execution across your entire infrastructure.