SECURITY ATLAS / EVALUATION BRIEF
Torq
Research reviewed: 2026-10-18
https://atlasofsecurity.com/landscape/vendors/torq/

THE ATLAS TAKE
Torq provides security workflow orchestration across integrations, secrets, and reusable automation. Atlas analysis: it is useful when a team wants to connect many operational steps while keeping the actual execution and dependency model inspectable. The hard part is governance of the accumulated authority. A workflow may access several systems through different credentials, and a reused integration can affect more processes than its author expects. Evaluate inventory, secret handling, execution evidence, and change ownership together. Claims about autonomous operations should be translated into a specific approved workflow with explicit human escalation and failure behavior.

SCOPE
This is licensed software the buyer configures and staffs. AI-assisted authoring still requires human control of go-live. Torq is not a managed detection and response service.

WORTH A LOOK WHEN
Teams building and operating cross-tool security automation with named workflow and integration owners.

THINK TWICE WHEN
The organization cannot inventory secrets and integration dependencies or treats autonomous workflows as a substitute for response policy.

IMPLEMENTATION
1. Define one bounded automation
Specify trigger, expected target, required API permissions, and analyst escalation.

2. Inventory shared resources
Record integration instances, secret owners, reuse, and rotation dependencies.

3. Test before promotion
Exercise duplicated input, denied permissions, and service failure with controlled lab targets.

OPERATING REALITY
Integrations, API keys, custom secrets, reusable workflows, and dependencies need recurring review. Atlas analysis: a credential rotation or shared step change can affect many paths. Keep a current resource inventory and test a controlled failure before allowing automation to make broad endpoint, identity, or cloud changes.

COMMERCIAL CONSIDERATIONS
Quote workflow capacity, governance, retention, integrations, and the operational support required for the intended program.
- Workflow and execution capacity
- Governance and operational feature entitlement
- Integration and shared-resource maintenance

EVALUATION PLAN / PROPOSED TESTS, NOT COMPLETED RESULTS
1. Trace a secret dependency
Acceptance evidence: Identify workflows using a lab secret and prove rotation affects only the intended integration.
Owner: __________  Pass criterion: __________
Observed result: __________  Evidence location: __________

2. Test an approval boundary
Acceptance evidence: Show a consequential lab action waits for the required decision and records the approver.
Owner: __________  Pass criterion: __________
Observed result: __________  Evidence location: __________

3. Audit a completed run
Acceptance evidence: Retain input, executed steps, target identity, result, and a visible failed-dependency path.
Owner: __________  Pass criterion: __________
Observed result: __________  Evidence location: __________

QUESTIONS FOR THE VENDOR
- Which workflow steps must stay human-gated regardless of HyperAgent suggestions?
- How are cases exported if the team later leaves Torq?
- Who is accountable when an AI-authored workflow takes an unauthorized infrastructure action, and how is that action rolled back?

EVIDENCE GAPS
- No independent secret-isolation or workflow-reliability assessment was performed.
- Public knowledge-base material does not establish every plan limit or execution option.

SOURCES
Torq custom secrets (Documentation)
https://kb.torq.io/en/articles/9142412-custom-secrets-protect-sensitive-data
Environment audit template (Documentation)
https://kb.torq.io/en/articles/11671901-workflow-template-what-do-we-have-environment-audit-report
Resource usage report (Documentation)
https://kb.torq.io/en/articles/11767202-workflow-template-where-used-resource-usage-report

Use synthetic fixtures in an authorized environment. Agree rollback and acceptance criteria before enforcement.
Documentation research and Atlas analysis. Independent hands-on effectiveness, effort, and complete price remain unassessed.