Close-Out Gate
A fix isn't finished until somebody can say what went wrong — and the answer is written down where it can be found again.
What it does
Collects every issue the tracker marks resolved or closed. Where a cause was recorded it restates it in one sentence; where none was, it proposes one from the evidence and says plainly that it is a proposal. It categorises each against a fixed taxonomy the analyst carries as knowledge, then asks a person to record it, send it back for a real cause, or mark it undeterminable. Confirmed causes are written to their own column — the memory that later work reads.
How it works
A fix is not finished until somebody can say what went wrong. This mission collects every issue the tracker marks resolved or closed, restates the cause where one was recorded and proposes one where none was, categorises it against a fixed taxonomy, and asks a person to confirm it. Confirmed causes are written down — so the next time something looks familiar, there is something to compare it to.
How it's wired
One poll watcher on the tracker's search, storing into `closeout`. Two enrichments by the RCA Analyst — a cause and a category, the latter grounded in the taxonomy document shipped on its bot — then one human gate whose three options each write back to the same row. The Close-out desk keeps proposals and confirmed facts in separate blocks.
The team
Turns a closed ticket into a stated cause — restating one that was recorded, or proposing one from the evidence and saying plainly that it is a proposal.
Connects to
What it watches
Polls the tracker every three minutes for issues marked resolved or closed, and records each one once.
Human approval
One consequential step waits for a person's approval before it acts.
What it already knows (4)
Data it carries
1 table
What it shows a human
Ask it
“State the root cause: "Refunds failed for orders created before the currency migration. PaymentMapper threw on a null currency."”
“Categorise this against the taxonomy: "Two concurrent requests each refreshed the session token; the later write invalidated the earlier one."”
“Here is a stated cause: "Fixed by adding a null check." Is that a cause? If not, what is missing?”
“What is the difference between a CONTRACT cause and a DATA cause?”
“When should a cause be recorded as UNKNOWN rather than guessed at?”
Before it runs, you'll
- 1Confirm the Issue Tracker server at import. It is credential-free and works as-is; replace the URL with your own tracker and nothing else changes — the mission only calls search_issues.
- 2Provision the RCA Analyst. Its root cause taxonomy uploads with it, and you can edit that document on the Knowledge tab without a redeploy.
- 3Enable the observation, then work the Close-out desk. Most tickets arrive with no cause recorded, so the first pass is where the memory gets built.
Import into StudioX
- Download the
.sxmfile. - In StudioX → Missions → Import, drop the file.
- Review, finalize, and it's live — chat to it or enable the observation.
Reviews
No reviews yet. If you’ve run this mission, yours will be the first — what it handled well, and what you had to wire up, is what the next person needs to know.
More in Engineering
Bug Intake Gate
Every bug report is read before it reaches the backlog — and the ones that can't be worked are named, not silently queued.
Triage Analyst
Defect Twin Finder
Before anyone starts work, the mission says whether this fault has been seen before — and what was learned the last time it was.
Signature Analyst
Change Impact Planner
Every commit is read as the files it actually changed — and what was decided about that area last time is on the screen when you scope this one.
Impact Analyst
