Run security assessments

Security tests lists discovery and assessment runs. Open a run to see its stage, statistics, selected scope, rule results and gaps. A run marked complete can still have a limited scope; read what it actually evaluated.

Passive assessments

Passive suites read supported source configuration. An Analyst, Admin or Owner can start them when the source is available.

1. Open Security tests and start a new assessment.
2. Choose an available passive suite.
3. Select the source integration to evaluate.
4. Review the scope and start the run.
5. Follow it through Queued, Discovering, Evaluating and Correlating.
6. Review the terminal result and any gaps before opening findings.

The exact suites and rules offered are returned by the application. Unavailable suites explain why they cannot be selected. Do not interpret a disabled suite as a hidden result.

Remote MCP authentication probe

This implemented safe-active suite sends a bounded MCP initialization request to selected remote servers to check whether an unauthenticated session is accepted. It does not enumerate and execute arbitrary tools or perform a complete authorization audit.

You need Admin or Owner and the active-assessments entitlement. The wizard requires explicit targets, a request budget and a written authorization statement. Only select systems you are authorized to test. The server validates and records the scope; it restricts private-address probing in production.

Local command-launched MCP servers are not equivalent to remote HTTPS targets and may be skipped with a coverage gap. A response from an endpoint can also depend on intermediaries, rate limits and the endpoint’s availability. Review evidence and gaps rather than relying on a single badge.

Completion states

State	Interpretation
Completed	The selected operation finished; review its statistics and supported scope.
Completed with gaps	Some requested work could not be fully evaluated. Follow its remediation.
Failed	The operation did not produce a successful complete evaluation. Inspect its error.
Cancelled	The cancellation was recorded; work already performed can still have produced evidence.

Use Cancel to stop further work on a run. Use Retry when the cause is resolved, reviewing its scope and authorization again where required. Build-to-build comparison and general model red teaming are not implemented.