Keeping the Product pivot (#18) means the coverage behind it has to be real. Today it
is not:
| Product |
Scenarios |
| Event Gateway |
14 |
| Outpost |
1 |
| Console |
0 |
The pivot renders a column per product, so Outpost's entire score rests on one
scenario and Console does not appear at all. A reader asking "how do agents do with
Console?" gets no answer, and one asking about Outpost gets an answer built on a
single measurement.
Console is the gap worth naming separately from #3. It is a distinct product surface
with its own API and its own failure modes, and at least one existing finding already
touches it: an agent retrying a request from the request page rather than the event
page, assuming delivery had been re-attempted when nothing was re-sent. That is a
Console navigation problem an agent surfaced, and there is no scenario covering it.
Scenario ideas, deliberately shaped like the ones that already discriminate, where an
agent can finish confidently and be wrong:
- Retry the right thing. Seed failed deliveries. Ask the agent to re-deliver them
through Console. Score whether the events were actually re-sent, not whether
something was clicked.
- Read delivery history correctly. Ask a question answerable only by navigating
Console's request and event views, where the two look similar and mean different
things.
Note Console's API is separate from the Event Gateway API the harness already speaks,
so this may need provisioning work before a scenario can be scored. Worth checking
that first, since it decides whether this is a day or a week.
Related: #3 for Outpost, #8 for the same problem on the stage axis.
Done when Console has enough scenarios to appear in the Product pivot honestly,
and Outpost's column is not a single run.
Keeping the Product pivot (#18) means the coverage behind it has to be real. Today it
is not:
The pivot renders a column per product, so Outpost's entire score rests on one
scenario and Console does not appear at all. A reader asking "how do agents do with
Console?" gets no answer, and one asking about Outpost gets an answer built on a
single measurement.
Console is the gap worth naming separately from #3. It is a distinct product surface
with its own API and its own failure modes, and at least one existing finding already
touches it: an agent retrying a request from the request page rather than the event
page, assuming delivery had been re-attempted when nothing was re-sent. That is a
Console navigation problem an agent surfaced, and there is no scenario covering it.
Scenario ideas, deliberately shaped like the ones that already discriminate, where an
agent can finish confidently and be wrong:
through Console. Score whether the events were actually re-sent, not whether
something was clicked.
Console's request and event views, where the two look similar and mean different
things.
Note Console's API is separate from the Event Gateway API the harness already speaks,
so this may need provisioning work before a scenario can be scored. Worth checking
that first, since it decides whether this is a day or a week.
Related: #3 for Outpost, #8 for the same problem on the stage axis.
Done when Console has enough scenarios to appear in the Product pivot honestly,
and Outpost's column is not a single run.