LabscoMCP early access ↗
REPO PACKAGEOFFICIAL

exploreomni omni-agent-skills

exploreomni · publisher40 GitHub starsgithub.com/exploreomni/omni-agent-skills

LABSCO SUMMARY

Summary

exploreomni/omni-agent-skills is Omni Analytics' own plugin set, and every skill in it needs a live Omni account and API key before it does anything.

Most ship in the omni-analytics plugin and walk a BI tool's usual lifecycle through the Omni CLI: omni-model-explorer and omni-model-builder read and edit the semantic model — views, topics, dimensions, measures, joins — in YAML; omni-query runs against it and chains multi-step analysis, including raw SQL and table calculations like running totals or period-over-period change; omni-content-explorer and omni-content-builder find, then build or edit, the dashboards and workbooks on top of that model; omni-embed signs URLs and wires themes and events for embedding a dashboard elsewhere; omni-admin covers users, groups, permissions, and schema refreshes; and omni-ai-optimizer and omni-ai-eval both concern Blobby, Omni's own AI agent — one tunes ai_context, ai_fields, and synonyms so Blobby answers correctly, the other runs judged evals to check whether it does.

The remaining two live in a separate omni-integrations plugin and each do one conversion: omni-to-snowflake-semantic-view and omni-to-databricks-metric-view turn an Omni topic into that warehouse's own semantic-layer format, for a team that wants the same metrics reachable from Snowflake Cortex Analyst or Databricks Genie.

READ MORE

Who this is not for.

There is no free or local path through any of them — the CLI needs an organization or personal API key before its first command runs, and several admin workflows specifically need an org-level key rather than a personal one. If your team doesn't already run Omni Analytics, none of these skills has anything to teach you.

One structural note.

The same repository also ships three Cursor .mdc rule files (API conventions, YAML conventions, terminology) that Cursor loads automatically but that Claude Code and other skills.sh agents do not load as rules at all — the README says as much, calling them reference material outside Cursor. We count skills, not rules, so those three sit outside the skills above.