openai publishes 1 MCP server and 19 agent skills on the atlas: 18 SAFE, 1 CAUTION, 1 BLOCK, with a mean score of 88/100. Every listing below is audited on its own.
Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
Cite
Cite this page
ModelsAtlas. “openai: skills and MCP servers”. Retrieved 28 Oct 2026. https://modelsatlas.com/orgs/openai
- Listings
- 20
- Mean score
- 88 / 100
- Findings
- 60
- Stars
- 569K
1 MCP servers · 19 skills
mean of the audit scores
across all listings
on the listed repositories
SAFE 18CAUTION 1BLOCK 1
Everything openai publishes here
Most stars first · stamps show the audited version
Questions
Are openai's MCP servers and skills safe to install?
Of 20 listings, 18 are SAFE, 1 CAUTION and 1 BLOCK, with a mean score of 88 out of 100. That is an average over separate audits: check each listing's own stamp before installing.
What does this page tell me about openai?
It collects every skill and MCP server on the atlas whose source lives under github.com/openai, with each one's verdict. It says nothing about who runs the organisation, and publisher identity is not verified here.
Can a publisher change its verdicts?
No. Each verdict comes from reading the source at a named commit. If a finding is wrong, report it with your reasoning on the listing's page and it is re-audited; the correction is public.
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- Are openai's MCP servers and skills safe to install?
- What does this page tell me about openai?
- Can a publisher change its verdicts?