Vulnerability evidence straight from the scanner
Replacing hand-exported scan files with fetchers that pull the report from the SentinelOne, Tenable, and Wiz APIs.
In most continuous monitoring programs, vulnerability data reaches the compliance platform the same way every month: someone logs into the scanner, exports a file, and uploads it. That works until the person is out, picks the wrong scan, or uploads last month's file twice. FedRAMP 20x expects vulnerability reporting to be persistent and machine-readable, and a hand-carried file is the weak link.
In paramify-fetchers, the open-source evidence repo I work on with the rest of the Paramify team, the answer is a kind of fetcher we call an issue-report fetcher. I've been building them for SentinelOne and Tenable. Teammates built the Wiz set.
Hand over the report, not a summary
Most fetchers in the repo collect configuration state and wrap it in a JSON envelope that validators read. An issue-report fetcher does something narrower. It pulls the report the scanner already produces, such as a Nessus export, a Wiz findings CSV, or a SentinelOne JSON, and writes it to disk byte for byte. There's no envelope and no added metadata. Paramify's assessment pipeline parses it into issues, the same way it would parse a manual upload.
That restraint is the point. The intake parsers already understand each scanner's native format. If the fetcher reshaped findings on the way through, every field I renamed or dropped would be a new place for the numbers to drift from what the scanner says.
Read-only, and scoped to the report
The credentials only need to read reports. Nothing in these fetchers writes to the scanner, and the access request names the scope the report needs rather than an admin role. That keeps the conversation with a client's security team short, which matters when the scanner belongs to a different team from the one running compliance.
An empty file is a failure
The most important rule is the least obvious one. If a vulnerability report comes back with no findings, the fetcher fails instead of uploading it. Intake reads "no findings" as "everything is resolved" and would close every open issue. A scan that never ran and a genuinely clean scan look the same as an empty file, so the fetcher refuses to guess.
The repo's template makes every issue-report fetcher state three caveats in its README: freshness (does the tool rescan on request, or return its last completed scan?), scope (what did the scanner never reach?), and how to tell an empty result from a clean one. Writing those down took longer than the API calls.
What changes for the client
The monthly export step goes away. The report that reaches the assessment is the one the scanner produced, pulled on a schedule, with a record of which run sent it. When an assessor asks where a finding came from, the answer is an API call rather than someone's downloads folder.