Explanation
Findings & severity
What a finding contains, where it comes from, how severity and confidence decide what is posted, and how the scores in the summary are computed.
A finding is one problem codingassist reports on a pull request: a severity, a file and line, a rule, a message and a confidence.
Where findings come from
| Source | Produced by | Confidence |
|---|---|---|
| Static analysis | The built-in rules of Stage 2 | 0.95 |
| Sandbox | The open-source analyzers run in the sandbox | 0.95 |
| Your tools | The sandboxTools of .praxiom.yml | 0.90 |
| AI review | The model call of Stage 3, after verification | Set by the model |
The AI review also reports duplicated code that the change introduces, with a confidence of 1.0.
What a finding contains
| Field | Content |
|---|---|
| Severity | Low, Medium, High or Critical |
| Location | File path and line |
| Rule | The rule or tool that raised it |
| Title and message | What is wrong and why |
| Confidence | 0 to 1 |
| Category | For AI findings: security, quality, acceptance criteria, or suggestion |
What gets posted
- AI findings with a confidence below 0.80 are dropped before publication.
- A finding with a file and a line, and a confidence of 0.80 or more, gets an inline comment.
- Every finding is counted in the summary comment.
Severity and the status check
The PRaxiom status check fails when at least one Critical finding remains, and passes otherwise. This threshold is fixed: .praxiom.yml cannot change it today.
A failed status check blocks a merge only if the target branch requires it. In Azure DevOps, add a status check to the branch policy and select the PRaxiom status.
Scores in the summary
| Score | How it is computed |
|---|---|
| Risk, 0 to 10 | The value returned by the AI review. If there is none: 4 per Critical, 2 per High and 1 per Medium finding, capped at 10 |
| Acceptance criteria match | Satisfied criteria divided by all criteria. Shown as n/a when no work item was found |
| Impact matrix | Security, Quality, Performance and Robustness start at 100 and lose 25 per Critical, 12 per High, 5 per Medium and 2 per Low finding of that kind. AC match is the match score as a percentage. Coverage is 60% AC match and 40% overall code health, or code health alone when there are no criteria. When the review was cut short, the axes are not scored |
Related
Was this page helpful?