SentinelOne package v1.1.0

Store the complete Authorization header value, ApiToken followed by one space and the console API token, in the customer's secret manager. The vendor-maintained ai-siem Management Console API reference documents that scheme and server-side permissions. An external deployment process must generate, store and rotate the token; this pack injects a stored value and does not manage its lifecycle.

Use a dedicated service user with only the endpoint permissions and account/site scope required by the enabled capabilities. The published references name Endpoints.view, Threats.view, Activity Page.view, Endpoints.disconnectFromNetwork and Endpoints.reconnectToNetwork, plus action-specific threat mitigation permissions. The old resolve route has no verified permission contract. Separate read and response credentials where the deployment requires that separation; this pack supplies only one reference.

The Service Users reference documents generation of a new service-user API token and revocation of the existing one. The Users reference also documents token revocation. The service-user listing includes apiTokenExpiresAt. These establish documented lifecycle operations and an expiry field, not a universal TTL, mandatory expiry, or immediate revocation timing on every console version. Confirm issuance options, expiry and cut-off behavior in the tenant before enabling. Treat captured material as live until those checks establish otherwise; refuse expired or rejected values.

The deployment must restrict vault reads to the required reference and keep the resolved header in executor memory only for the call, out of logs, traces, diagnostics and crash reports. These controls and the remaining lifecycle questions are pending owner.

Draft: the console API Hub remains the tenant-specific reference. On 2026-10-29 the public Sentinel-One/ai-siem repository supplied v2.1 operation references for agents, threats, activities, mitigation, disconnect and connect. Its README says SentinelOne maintains it with community support and that it is not a formal SentinelOne product. This is vendor-published reference material, with that support boundary; it does not establish compatibility with every tenant version.

The old POST /web/api/v2.1/threats/mark-as-resolved registration remains UNVERIFIED. The published Threats reference does not list that route. It lists a separate /threats/incident operation, but the full body schema was not retrieved, so this pack does not substitute or claim that operation is equivalent. Do not enable the resolve capability. The complete Swagger file exceeded the web reader's size limit, leaving nested request/response schemas unverified. The pack remains HELD.

Replace sentinelone-console.invalid with the reviewed customer console host in every endpoint, resource locator and egress allowlist together. Keep ACTION unresolved until a concrete permitted mitigation has been reviewed. The published action enum uses rollback-remediation for rollback. Its prose and enum differ on other selectors; refuse uncertain selectors until the tenant API Hub resolves the difference. This pack does not restrict request bodies or ACTION values by itself.

Activities can identify the user who invoked an action; the reference exposes userIds and userEmails filters. The activities capability and its console resource are classified as pii. Agents remain pii; threat and response registrations remain internal, subject to full response-schema and tenant data review. These are AAES handling judgments. Reads declare read/R0; resolve declares update/R2; isolation and mitigation declare deploy/R3; reconnect declares update+deploy/R3. These tiers are action floors, not provider risk ratings.

The pack registers no agent uninstall, decommission, remote shell, RemoteOps, scans, policy/exclusion/blocklist changes, account/site/user/role administration or ranger configuration. The prior description grouped decommission with irreversible removal of protection; the published Agent Actions reference instead describes removal from the console and automatic recommission on later agent communication. Both operations stay excluded. Standalone rollback is excluded, but mitigation can still request rollback-remediation unless the deployment restricts ACTION. Other mitigation selectors are not claimed absent from the provider API.

A reviewer must finish the unresolved route, nested schemas and credential lifecycle checks before promotion. Token scope, expiry and revocation checks are pending owner. No owner approval or tenant execution is established by this record.