Code-to-exploit stages
A critical vulnerability in a package nobody imports is a chore. The same vulnerability in a service on the public internet is an incident. Severity alone calls those the same thing.
The code-to-exploit stages tell them apart by asking how far a problem has travelled toward an attacker.
| Stage | What it means | What makes it known |
|---|---|---|
| Present | The problem exists in code. | Code |
| Reachable | Your code can actually reach it. | Code |
| Shipped | It is inside an artifact you build. | Artifacts |
| Running | That artifact is running in your cloud. | Cloud & Workloads |
| Exposed | It can be reached from the internet. | APIs |
| Proven | Someone showed it can be exploited. | APIs |
Where this stands today
Findings do not carry a stage yet, so the Stage column on the findings list reads Unknown on every row. For now, rank by severity plus the EPSS and KEV signals described in Findings.
Unknown is not the same as safe
Here is the part people get wrong. If only GitHub is connected, Dwaar has no way to see where anything is running, so a finding cannot move past Reachable. That finding is not "not shipped". It is unknown.
The console draws unknown stages dashed, with a link to the integration that would fill them in. Treat each one as a question you have not answered yet.
Try this next
Connect the next integration along the chain now, so its data is already there when stages arrive on your findings. Integrations shows what each one adds.