Barua can review recorded activity and suggest a useful email Play when the evidence supports one. Each suggestion shows the documented help it would offer. You review the complete emails before approving a Play.
Read the evidence
Use BARUA_MANAGEMENT_KEY with the installed skill:
node "$BARUA_SKILL/scripts/manage.mjs" capabilities
node "$BARUA_SKILL/scripts/manage.mjs" discovery
Follow a returned next cursor with discovery --after. Read one result with
discovery-opportunity --id, using its returned ID. The same evidence appears
under Opportunities in Plays.
The status distinguishes a review that has not run from one that failed or completed. A completed review may find nothing useful to propose. Partial coverage means some activity is missing; it does not mean no activity occurred. The discovery contract defines every field and action.
Understand the scope
Each opportunity names its evidence window and sample. Pixel sessions and server occurrences remain separate counts. A lower-bound count means at least that many observations. Samples include extra task-related activity, so their mixture does not measure the problem's frequency across all visitors.
Current eligibility applies only within the sample. Missing email permission, suppression or old activity can prevent sending. A captured address alone is insufficient. Setup blockers remain visible alongside the evidence, including an unapproved tracking plan or missing completion event.
Review or refresh
Discovery changes require play:write. Use the management action command
with the exact current tokens described in the reference. A refresh requests
a rescan within the usual daily limits. It preserves existing proposals and
dismissals. Replacing an unapproved proposal is a separate explicit decision;
read its replacement context first.
Snoozing a suggestion also defers its pending preparation. After the snooze ends, Barua checks the saved work again under the usual freshness and preparation limits.
Discovery can be paused without changing approved Plays. Enabling discovery does not enable pixel execution or sending.
When a suggestion has a linked Play, inspect its rules and complete emails. Use the ordinary exact-revision approval procedure. A prepared proposal still needs approval. Historical observations cannot manufacture fresh events, while approval may admit activity that remains inside the live eligibility window. Before approval, discovery proposals keep their fixed rules and do not offer Request changes. Discard an unwanted proposal or use its available replacement action. After adoption, ordinary Play changes and language review are available; expired discovery evidence does not revoke an already approved Play.
Read later results
Results show sends and measured goal outcomes for the returned window. Fewer than 20 measurable outcomes is insufficient to judge effectiveness. Unmeasured sends have no known goal result. Completion after an email does not prove the email caused it. Discovery can suggest another review; it does not silently edit an approved Play.