Repository navigation
feat(evals): score Harbor skill evidence in the example - #4171
Merged
ngoncharenko merged 1 commit intoNov 19, 2026
Merged
ngoncharenko merged 1 commit into
ngoncharenko merged 1 commit into
Conversation
Signed-off-by: abondarenko <abondarenko@nvidia.com>
ngoncharenko
approved these changes
Nov 19, 2026
ngoncharenko
merged commit Nov 19, 2026
05b0cad
into
harbor-evals-skill-example/ngoncharenko
4 checks passed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The Harbor example now scores package readiness from the verified skill snapshot and command outcomes from saved Codex trials. Replay carries outcome flags for matched root
exec_commandcalls and marks the trace partial. Unknown outcomes are counted separately from successes and failures.Run
score_evidence_example.pyto see synthetic trace, scan, and bundled-test evidence and their scores. The fixture runs offline and is separate from live scoring.Stacked on #4164.
Validation includes 573 targeted SDK, Harbor runtime, and plugin tests, plus the offline fixture. The full local hook run stopped on unavailable documentation tooling. Live Harbor execution was not part of this check.