Check Minecraft Plugin Compat

check-minecraft-plugin-compat

Check whether plugins pinned via direct CDN URL (because they lacked a tagged Paper build for the running Minecraft version) have since published real support, and whether plugins marked optional (?) in MODRINTH_PROJECTS can be re-enabled. Reports findings and offers to apply safe reverts -- does not auto-merge anything.

Background

During the 26.2 Paper upgrade (see git history around commit af4d1bb57 and surrounding commits in kubernetes/default/minecraft/), several plugins had no Paper build tagged for 26.2 in their Modrinth/Hangar metadata. Two different fixes were applied depending on the actual root cause, which this skill must distinguish between:

• Metadata-only gap (plugin jar actually works, just not tagged): pinned via APPLY_EXTRA_FILES: plugins<<CDN-URL>> to bypass the Modrinth/Hangar compatibility resolver entirely. Example: essentialsx, locked-chests-plugin, luckperms.
• Real incompatibility (plugin code itself refuses to run / crashes on the new MC version): marked optional in MODRINTH_PROJECTS (coreprotect?) so it's cleanly skipped instead of installed-but-broken. nochatreports-spigot-paper was outright replaced with a different plugin (BlockReports) because its NMS-version lookup crashed.

Critical lesson learned: do not trust Modrinth's declared game_versions metadata or even an explicit project:versionId pin as proof a plugin works. CoreProtect's jar downloads fine via direct URL but prints Minecraft 26.2 is not supported. and disables itself at runtime. NoChatReports crashes with a NullPointerException in its own NMS version-detection code. Always check the actual pod startup logs after a test restart, not just whether the resolver/download succeeded.

There is no Renovate datasource (native or community) for Modrinth/Hangar plugin versions -- confirmed via web search, this is a known gap. That's why this is a skill instead of a .renovate/*.json5 custom datasource: Modrinth CDN URLs embed an opaque version ID and a filename that must change together, and there's no reliable way for Renovate's versioning schemes to determine which opaque ID is "newer."

Usage

/check-minecraft-plugin-compat

No arguments. Scans all three server configs under kubernetes/default/minecraft/.

Steps

1 -- Find the current pins and optional entries

rg -n "plugins<https://(cdn\.modrinth\.com|hangarcdn\.papermc\.io)" kubernetes/default/minecraft/*.yaml
rg -n -A1 "MODRINTH_PROJECTS" kubernetes/default/minecraft/*.yaml

From the APPLY_EXTRA_FILES URLs, extract the Modrinth project ID (segment after /data/) or Hangar owner/slug (segment after /plugins/). From MODRINTH_PROJECTS, note any entries ending in ? -- those are the optional/skipped ones to re-check.

2 -- Determine the running Minecraft version

Read it from the live world data rather than assuming -- the cluster may have moved on since this skill was written:

kubectl exec -n default minecraft-survival-0 -- cat /data/version_history.json

Use the currentVersion MC number (e.g. 26.2) as the target to check compatibility against.

3 -- Query Modrinth for each pinned/optional project

curl -s "https://api.modrinth.com/v2/project/<slug-or-id>/version?loaders=%5B%22paper%22%5D" \
  | jq -r '.[0] | "version_number=\(.version_number) id=\(.id) date=\(.date_published) game_versions=\(.game_versions[-5:])"'

A project is a candidate for reverting to plain auto-resolve (no pin, no ?) only if the target MC version string appears in game_versions. A metadata match is necessary but not sufficient -- proceed to step 5 before declaring it fixed.

4 -- Query Hangar for Hangar-hosted plugins (e.g. BlockReports)

curl -s "https://hangar.papermc.io/api/v1/projects/<owner>/<slug>/versions?limit=5" \
  | jq -r '.result[] | "\(.name) platforms=\(.platformDependencies)"'
curl -s "https://hangar.papermc.io/api/v1/projects/<owner>/<slug>/versions/<version>" \
  | jq -r '.platformDependencies'

Also worth periodically re-checking whether nochatreports-spigot-paper itself has published real 26.2 support, in case the team wants to revert from BlockReports back to it -- check the actual crash signature is gone (see step 5), not just the version tag.

5 -- Functionally verify candidates before recommending a revert

For each candidate found in steps 3-4, this skill must not just report "metadata now matches" -- it must actually test it, the same way this was diagnosed originally:

1. Edit the relevant MODRINTH_PROJECTS/APPLY_EXTRA_FILES entry locally (uncommitted).
2. flux suspend helmrelease <app> -n default to stop Flux from fighting a failed test.
3. Push a throwaway commit is NOT needed for this -- instead patch the live StatefulSet pod env directly is also not appropriate (see GitOps note below). Prefer: commit the candidate change to a scratch branch, but do not push/reconcile against master until step 5 confirms it. If a live functional test is needed, ask the user first -- direct pod/PVC mutation outside git is a deliberate, confirmed exception, not a default action (this was flagged mid-session previously).
4. Check kubectl logs -n default <pod> | rg -i "error|exception|disab|not supported" for the specific plugin's enable sequence. Only a clean Enabling <Plugin> with no following ERROR/Exception counts as confirmed working.

6 -- Report findings

Plugin              Pin type        Latest available    Target MC tagged?   Functionally verified?   Recommendation
essentialsx         CDN pin         2.23.0               yes                 not yet tested            test before reverting
coreprotect         optional (?)    23.3                  no                  n/a                       still skip
nochatreports       (replaced)      2.8.0                 yes                 not yet tested            consider reverting from BlockReports

Ask the user before applying any revert -- do not auto-commit. If they approve, follow the same careful rollout procedure used originally: suspend the HelmRelease, commit + push + reconcile, resume, watch the pod via Monitor until ready, then check logs for the specific plugin's enable line before declaring success.