JevMade Sign in
← Back to videos

JevMade field notes / Video guide

Jev Agent Harness Tutorial: Skill Selection, RAG Re-ranking and Known Limits

Pico AI Lab demonstrates choosing a skill in two steps and ranking retrieved passages, with fallbacks and limits for each decision.

Original by Pico AI LabAgent workflowsIntermediate10 min 20 sec Published

Before you press play

What you’ll find in the video

  1. Choose a category, then a skill within it. Include an explicit no-match option and fall back when the result is uncertain.
  2. Use a clear rating scale to reorder retrieved passages before sending a smaller selection to the language model.
  3. Keep counting, arithmetic and date comparisons in code. Screening text for injected instructions is an extra probabilistic check, not a security boundary.
Worth knowing

Toy demonstrations with no runnable implementation linked in the publisher description. Speed, cost and test results are reported by the narrator and were not reproduced. The example injection cutoff does not establish safety.

Jev Agent Harness Tutorial: Skill Selection, RAG Re-ranking and Known Limits

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.