Catalog checks: what we found in the listed entries
Last updated 7 November 2026
This report covers the entries that Spellwarden chose to list in its catalog after review: 52 of them. It is not a survey of the public ecosystem and not a view of the entries that Spellwarden declined, so no number here is a rate across all skills or servers. Data as of 6 November 2026, from catalog export of scanner 1.8.1 with rules 2026.11.1.
Entries checked
52 entries are listed: 37 skills, 14 MCP servers and one plugin.
| Category | Entries |
|---|---|
| Cloud platforms | 12 |
| Infrastructure as code | 7 |
| Kubernetes and containers | 7 |
| Pipelines and delivery | 4 |
| Observability | 15 |
| Security operations | 6 |
| Documentation servers | 1 |
Skill packages and curation scores
The listed skills and plugins hold 42 skill packages; 42 of them are Curated.
42 packages have a curation score, on a scale of 0 to 12. The lowest is 8.0 of 12, the median 9.2 of 12 and the highest 10.8 of 12.
Open findings by severity
Across the listed entries the open findings are 0 critical, 0 high, 19 medium and 37 low. Open findings are the recorded findings without the ones the catalog accepted.
No listed entry has an open critical or high finding.
Accepted findings by rule
The public catalog file records the findings of each check as counts per severity, which include the accepted findings, and a rule id only for findings the catalog accepted, so this table lists those. It does not show how often each rule fires.
| Rule | Accepted findings | Severities |
|---|---|---|
| SW-MCP-001 | 1 | high |
Method
Each entry is scanned with the open-source scanner's rules, reviewed, and checked again when it changes; the catalog guide describes the checks. The report names no publisher and no entry and ranks none.
Browse the entries in the catalog, and read how the attacks behind the rules work in MCP tool poisoning and Prompt injection in agent skills.