How ATLAS checks itself

Every check the project runs on its own pipeline, collected in one place: whether the Electoral Division join reproduces against live sources, how the sources themselves score, whether the cartography rules still hold, how the narrative gate treats a model's text, and what the test suite found the last time this copy was built. None of it rates a place.

This summary was built:

The Electoral Division join, checked against live sources

Reproduced against live sources

What that join resolved to in this copy's data

This reproduces a join, not the correctness of the underlying census or boundary data, and it is not a rating of any Electoral Division. A GUID with no matching name is a finding about the source files, not about the place.

Source scorecard

Some detail on this page is generated automatically by ATLAS's own checks, in English only, and has not been translated.

Source scorecard: retrieval, licence, schema and join safety for each source
Source	Overall	Last retrieval	Licence	Schema	Join safety

PASS, WARN and FAIL describe whether a source behaved as expected on the date shown. They are not a judgement of the publisher, and WARN on the join-safety column is a known, worked-around limitation (see the join checks above), not a live fault.

Cartography rules

Some detail on this page is generated automatically by ATLAS's own checks, in English only, and has not been translated.

The cartography rule checks, their result and detail
Rule	Check	Result	Detail

A check passing means the interface behaved as the rule requires on the date shown; it says nothing about whether a rule was the right one to write, and it does not rate any Electoral Division or council.

The narrative gate

The narrative gate's pass rate for each writer it has evaluated
Producer	Cases	Passed	Pass rate	Model text accepted

Some detail on this page is generated automatically by ATLAS's own checks, in English only, and has not been translated.

The test suite

A green suite means the code behaved as its own tests expect; it is not a claim that the tests cover everything worth testing.

What this does not show

None of this rates or scores any place, community or council, and it is not a ranking of any kind. It rates whether the pipeline's own checks passed, never the data itself or the areas it describes.
A check that passes says the code behaved as written, on the date shown. It is not a claim that the underlying public data is accurate, complete or current — see the Sources page for that.
Only what a check itself wrote is left untranslated, even in Irish mode: the cartography rule and detail text, each source's licence/schema/join-safety detail, and the model's rejected text. Labels, dates and connecting words around them are ATLAS's own wording and are translated.

Per-dataset licence, publisher and retrieval age are on the Sources page, not here.