1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
23 small wins to finish your pathNext lesson →
Generators that need no key
Before pointing a scan at anything that costs money, it is worth knowing which targets are free, because they are how you check a scan is set up correctly.
| Generator | What it answers | What it is for |
|---|---|---|
| test.Blank | an empty string | checking a probe runs at all |
| test.Repeat | the prompt it was given | seeing what a probe actually sends |
| test.Lipsum | filler text | giving detectors something to chew on |
| test.Single | one fixed string | a fixed answer for a detector test |
| function | whatever your Python returns | your own application |
test.Repeat is the useful one. Point a probe at it and the report holds exactly what that probe sends, which is how lesson 10 reads a probe without reading its source.
garak -t test -n Repeat --spec probes.test.Blank --generations 1 --report_prefix echoThe run passes, because an empty prompt repeated back is still empty and the detector has nothing to report. What matters is the report it wrote, not the verdict.
These are not a substitute for a real target. They exist so that a scan which fails tells you something. If a probe errors against
test.Blank the problem is the probe or the install; if it only fails against your function, the problem is yours.Try it yourself
- Run the same probe against
-n Lipsumand compare the verdicts. - Run it against
-n Singleand find where the fixed answer is configured. - Work out which of the five you would use to test a detector you just wrote.
Every expert started right here.