The README tells readers how to get a stable snapshot:
Pin to a commit instead of @main if you need a stable snapshot.
There is nothing else to pin to — the repository has 0 tags and 0 releases. A commit SHA works, but it says nothing about what is in it: a consumer pinning today cannot tell whether they are on the refresh that fixed hero prices or the one that broke them, and there is no list of what changed between two pins.
The weekly job is what makes this matter. main moves on its own every Monday, so "pin if you need stability" is the honest advice — it just has no good target.
Suggested fix
Tag each refresh with its date and let the workflow do it, since the workflow is what creates the change:
2026-08-24, 2026-08-31, …
https://cdn.jsdelivr.net/gh/MLuc24/lien-quan-data@2026-08-24/data/heroes.json then means something to someone reading it a year later, and the tag list becomes the changelog nobody has to write. A GitHub release per tag with the record counts of that snapshot would be the nicer version, and the counts are already printed by the merge step.
Related: there is no cheap way to read a little
data/heroes.json is 1.5 MB. The most common thing to build first — a search box, a dropdown, an index page — needs the slug, the name, the classes and a portrait, and today that costs the full download of every skill description, every skin URL, every balance-history line for all 129 heroes.
A generated data/index.json carrying just those fields would be a few tens of KB. It is derived, so it belongs in the merge step rather than being maintained by hand, and it costs nothing to keep correct.
Both of these are additive — no existing path changes, so nothing anyone already fetches breaks.
The README tells readers how to get a stable snapshot:
There is nothing else to pin to — the repository has 0 tags and 0 releases. A commit SHA works, but it says nothing about what is in it: a consumer pinning today cannot tell whether they are on the refresh that fixed hero prices or the one that broke them, and there is no list of what changed between two pins.
The weekly job is what makes this matter.
mainmoves on its own every Monday, so "pin if you need stability" is the honest advice — it just has no good target.Suggested fix
Tag each refresh with its date and let the workflow do it, since the workflow is what creates the change:
https://cdn.jsdelivr.net/gh/MLuc24/lien-quan-data@2026-08-24/data/heroes.jsonthen means something to someone reading it a year later, and the tag list becomes the changelog nobody has to write. A GitHub release per tag with the record counts of that snapshot would be the nicer version, and the counts are already printed by the merge step.Related: there is no cheap way to read a little
data/heroes.jsonis 1.5 MB. The most common thing to build first — a search box, a dropdown, an index page — needs the slug, the name, the classes and a portrait, and today that costs the full download of every skill description, every skin URL, every balance-history line for all 129 heroes.A generated
data/index.jsoncarrying just those fields would be a few tens of KB. It is derived, so it belongs in the merge step rather than being maintained by hand, and it costs nothing to keep correct.Both of these are additive — no existing path changes, so nothing anyone already fetches breaks.