Skip to main content
audit runs a feature-by-feature or group-by-group review using ForEachFeature. It is designed to consume the output of feature-enum, but you can provide any feature map: an object whose keys are group names and whose values are arrays of feature identifiers: 
{ "GROUP_NAME": ["FEATURE_A", "FEATURE_B"] }

bunx smithers-orchestrator workflow run audit \
  --input '{"features":{"GATEWAY":["GATEWAY_RPC_LIST_RUNS"]},"focus":"test coverage"}'

​
Inputs

InputTypeDefault
featuresobject{}
focusstring"code review"
additionalContextstring or nullnull
maxConcurrencynumber5

​
Output

The workflow returns two structured outputs in the run result (nothing is written to disk):
KeyShapeDescription
auditFeature{ groupName, result, featuresCovered[], score? }One entry per group processed in parallel
audit{ totalGroups, summary, mergedResult, markdownBody }Merged report across all groups

​
Use it when

  • You have a feature map (from feature-enum or your own) and want a gap review per group.
  • You want findings on missing tests, docs, observability, error handling, or maintainability work.

​
Use something else when

  • You do not have a feature inventory yet: run feature-enum first.
  • You want to review a diff rather than feature groups: use review.

​
Resume & retry

audit processes groups in parallel and returns structured results in the run output; nothing is written to disk. Runs are durable and resume with up --resume.