cleandataengineering.aiAgent Harness, drawn simply
You are here: End to end › Checks › 2 · Before the model › Prompt Guard 2

Agent Harness Tutorial · End to end · Zoom level 3

How we take care of it: Prompt Guard 2

Prompt Guard 2 catches paraphrased and foreign-language attacks.

Open: 11before loopopen ›Open: 22before modelopen ›Open: 33after modelopen ›Open: 44before toolopen ›Open: 55the toolopen ›Open: 66after toolopen ›Open: 77after loopopen ›Open: 88before replyopen ›It readsthe user message, and later eachtool resultPrompt Guard 2small model · open sourcepassCarries onfailIf it failssafe reply / hold writesWHERE IN THE CODE: planned · finding F-132
The strip at the top shows every check point where this check runs, in dark.

How we took care of it.

  1. Not built yet. Llama Prompt Guard 2 is a small open classifier for injection attempts.
  2. It would plug in as an hf_classifier evaluator: one line in evaluators.yaml, no new port.
  3. It would catch what the phrase list misses: paraphrases and other languages.
It readsthe user message, and later each tool result
Kindsmall model
Provided byopen source
If it failssafe reply / hold writes
Where in the codeplanned · finding F-132
Runs at2 · Before the model, 6 · After the tool
Real example—