How it works¶
The flow¶
targets ──► target.Parse ──► Nmap ──► hosts + services ──► checks ──► findings ──► report
(validate) (nmap) (engine) (output)
- Validate targets. We reject anything that isn't an IP, network, range or hostname. This also stops tricks like passing
-oN fileas a target. - Find hosts with Nmap. We run
nmap -sT -sV --openand read its XML. Nmap does all the low-level packet work. - Pick checks. Each check says which services it cares about (by port or Nmap service name). We only run a check where it applies.
- Run checks in parallel. Each check gets its own timeout. If a check errors, hangs or even panics, it's recorded and the scan keeps going.
- Report. Findings are sorted worst first, with evidence and a fix.
Where the code lives¶
See Architecture. It shows every package, what it does, what it uses, and every file in the repo. It's generated from the code on every docs build, so it's always right.
Rules we keep¶
- Simple first. If it can be simple, it should be simple.
- No hidden magic. Checks are listed by hand in
checks/checks.go, there's no auto-registration. - Standard library only for now, zero dependencies. Less to audit, less to break.
- Errors aren't findings. A check that couldn't connect never reports "vulnerable".
- Evidence for everything. Every finding says what we saw, so people can trust it.
- Confidence for everything. Every finding is
confirmed(we saw it happen) orinferred(only from a version number, or Nmap guessed the service from the port). - Same input, same output. Reports are sorted the same way every time.