Omni Analytics Agent Skills

Official Omni Analytics skills for Claude Code, Cursor, OpenAI Codex, Snowflake Cortex Code, and other skills.sh compatible agents.

Bring governed Omni workflows directly into your agent environment with one install target for model exploration, querying, dashboard creation, semantic modeling, AI optimization, administration, and embed work.

How It Works

These skills are not slash commands or one-off prompts you have to memorize. Once installed, the agent loads the relevant skill or agent when your request matches the use case.

• Ask in natural language for modeling, querying, dashboard, admin, or embed work
• Claude Code and Cursor can auto-load the right skills and agents
• Cursor can also apply the .mdc rules in this repo when relevant
• OpenAI Codex, Gemini CLI, and GitHub Copilot can load the same skills through skills.sh
• Snowflake Cortex Code can load the repo's SKILL.md directories as custom skills

Why This Repo

• One shared source of truth for Omni workflows across major agent surfaces
• 9 production-focused skills covering exploration, querying, modeling, content, admin, AI optimization, AI eval, and embed
• 3 specialized agents for deeper multi-step work
• 4 Cursor rules for API, YAML, terminology, and permissions consistency

Platform Support

Platform	Install method	Notes
Claude Code	Plugin marketplace or Git URL	Full plugin install with 9 skills and 3 agents
Cursor	Git URL plugin install	Full plugin install plus 4 .mdc rules
OpenAI Codex	npx skills add	Uses the shared skills.sh install flow
GitHub Copilot	npx skills add	Uses the shared skills.sh install flow
Gemini CLI	npx skills add	Uses the shared skills.sh install flow
Snowflake Cortex Code	Upload skill folders or copy into .cortex/skills/	Loads the repo's SKILL.md directories as custom skills
Other skills.sh-compatible agents	npx skills add	Same install flow as Codex, Copilot, and Gemini CLI

Installation

Claude Code

Marketplace install (recommended, run separately):

/plugin marketplace add exploreomni/omni-agent-skills

/plugin install omni-analytics@omni-analytics

To also install the integrations plugin (Snowflake Semantic Views, etc.):

/plugin install omni-integrations@omni-analytics

Git URL install (run separately):

/plugin marketplace add https://github.com/exploreomni/omni-agent-skills.git

/plugin install omni-analytics@omni-analytics

To also install the integrations plugin:

/plugin install omni-integrations@omni-analytics

Cursor

Install from Git URL:

/add-plugin https://github.com/exploreomni/omni-agent-skills.git

To also install the integrations plugin (Snowflake Semantic Views, etc.), use Cursor's team marketplace import (Dashboard > Settings > Plugins > Import) with the repo URL https://github.com/exploreomni/omni-agent-skills — both omni-analytics and omni-integrations will appear as separate installable plugins.

Cortex Code

Cortex Code can load these skills directly as custom skill folders. For the CLI, copy one or more folders from this repo's skills/ directory into project-local .cortex/skills/ or user-level ~/.snowflake/cortex/skills/. In Snowsight workspaces, you can upload the same skill folders directly.

Install all repo skills in the current project:

mkdir -p .cortex/skills
cp -R skills/* .cortex/skills/

Install one skill directly:

mkdir -p .cortex/skills
cp -R skills/omni-query .cortex/skills/

Install the integrations skills (Snowflake Semantic Views, etc.):

mkdir -p .cortex/skills
cp -R skills/omni-integrations .cortex/skills/

skills.sh-Compatible Agents

Use this path for agents that load skills through skills.sh, including OpenAI Codex, Gemini CLI, GitHub Copilot, and other compatible agent runtimes.

Preview available skills:

npx skills add exploreomni/omni-agent-skills --list

Install the full collection:

npx skills add exploreomni/omni-agent-skills

Install one skill directly:

npx skills add https://github.com/exploreomni/omni-agent-skills --skill omni-query

Install the integrations skills (Snowflake Semantic Views, etc.):

npx skills add https://github.com/exploreomni/omni-agent-skills --skill omni-integrations

Install globally:

npx skills add exploreomni/omni-agent-skills --global

Check for updates:

npx skills check

Update installed skills:

npx skills update

Migrating from Older Repos

This repo replaces two earlier, platform-specific repos that are now deprecated:

Deprecated repo	Replacement
exploreomni/omni-claude-skills	Use the Claude Code install above
exploreomni/omni-cursor-plugin	Use the Cursor install above

If you previously installed from either repo, uninstall the old plugin first, then follow the installation instructions above.

Claude Code

/plugin uninstall omni-analytics@omni-analytics
/plugin marketplace remove omni-claude-skills

Cursor

/remove-plugin omni-cursor-plugin

Then follow the installation instructions above to install from this repo.

Setup

Install the Omni CLI

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/exploreomni/cli/main/install.sh | sh

This downloads the latest release, verifies the SHA-256 checksum, and installs the omni binary to /usr/local/bin (or ~/.local/bin if not writable). Pre-built binaries for all platforms are also available on the GitHub Releases page.

Configure authentication

# Show available profiles and select the appropriate one
omni config show
# If multiple profiles exist, pick one and switch:
omni config use <profile-name>

If no profiles exist, run omni config init to create one interactively. You can also use --profile, --base-url, or --token flags to override the active profile for a single command.

API keys are created in Settings > API Keys (Organization Admin) or User Profile > Manage Account > Generate Token (Personal Access Token).

Admin note: Some workflows, especially schema refresh, permissions, schedules, and other admin operations, require an Organization API Key.

What You Get

omni-analytics — Skills (9)

These activate from natural-language requests:

Skill	Description
omni-model-explorer	Discover and inspect models, topics, views, fields, dimensions, measures, and relationships
omni-model-builder	Create and edit views, topics, dimensions, measures, and relationships in YAML
omni-query	Run queries against Omni's semantic layer and interpret results
omni-content-explorer	Find, browse, and organize dashboards, workbooks, and folders
omni-content-builder	Create, update, and manage documents and dashboards programmatically - lifecycle, tiles, filters, layouts
omni-ai-optimizer	Optimize your Omni model for Blobby, the Omni Agent
omni-admin	Manage connections, users, groups, permissions, schedules, and schema refreshes
omni-embed	Embed Omni dashboards in external applications - URL signing, themes, and postMessage events
omni-ai-eval	Evaluate AI query generation accuracy — run test prompts, compare results, and score across dimensions

omni-integrations — Skills (3)

Skill	Description
omni-to-snowflake-semantic-view	Convert an Omni Analytics topic into a Snowflake Semantic View YAML definition
omni-to-databricks-metric-view	Convert an Omni Analytics topic into a Databricks Metric View definition
omni-to-dbt-metricflow	Move Omni view and relationship logic into dbt MetricFlow YAML, then make Omni fall back to the dbt definition

Agents (3)

These are built for heavier workflows where explicit delegation is useful:

Agent	Description
omni-analyst	Explores models, runs queries, and delivers insights
omni-modeler	Builds semantic models, writes YAML, and optimizes for AI
omni-admin-agent	Manages users, permissions, schedules, and connections

Cursor Rules (4)

These .mdc files are included for Cursor's rules system:

Rule	Description
omni-api-conventions	Auth headers, base URL patterns, error handling, and pagination
omni-yaml-conventions	YAML file types, field syntax, and dimension and measure patterns
omni-terminology	Maps business intelligence terms to Omni-specific vocabulary
omni-permissions	Determine a caller's access via whoami (branch vs workbook) and assign model roles

Platform note: Cursor loads the rules/ directory automatically. Claude Code, Cortex Code, and skills.sh-compatible agents do not load these files as plugin rules, so outside Cursor they should be treated as shared reference material unless you copy them into that tool's own rule mechanism.

Example Prompts

Ask naturally:

"What topics are available in our Omni model?"
"Run a query showing revenue by month"
"Add a new dimension for customer tier to the users view"
"Find the dashboard about sales performance"
"Build a dashboard showing revenue by month"
"Improve the AI context on our orders topic"
"Give the marketing team access to the sales dashboard"
"Generate a signed embed URL for this dashboard"
"Convert this Omni topic to a Snowflake Semantic View"
"Move the measures on this Omni view into dbt semantic models and metrics"

For direct agent routing:

@omni-analyst What are our top 10 products by revenue this quarter?
@omni-modeler Add customer lifetime value metrics to the users view
@omni-admin-agent Set up weekly PDF delivery of the executive dashboard

Team Deployment (Claude Code)

To make this plugin available to your entire team automatically, add it to your project's .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "omni-analytics": {
      "source": {
        "source": "github",
        "repo": "exploreomni/omni-agent-skills"
      }
    }
  },
  "enabledPlugins": ["omni-analytics@omni-analytics", "omni-integrations@omni-analytics"]
}

When team members trust the repository folder, Claude Code automatically installs the marketplace and plugin.

Updating

Claude Code

Enable auto-updates:

1. Run /plugin
2. Go to Marketplaces
3. Select the marketplace → Enable auto-update

Or update manually from the /plugin menu.

Cursor / Cortex Code / skills.sh-Compatible Agents

Re-run the installation command or re-copy the updated skills/ folders to pull the latest version.

How These Skills Track the Omni CLI

These skills document a moving target: the Omni CLI ships new commands, and occasionally removes them. When a CLI release changes the command surface, a maintenance agent re-checks the skills against the newly released binary and opens a sync PR — so an updated plugin describes the CLI you actually have.

Sync releases are recorded in CHANGELOG.md, naming the CLI version each one covers. If you are running a newer CLI than the most recent sync, some commands may not be documented here yet; omni <command> --help and --schema are authoritative in that gap. omni update check reports whether your CLI itself is current.

Getting the Latest Version

If your skills appear outdated, or the install command reports the plugin is "already installed" while skills aren't loading, your IDE may be serving a cached version. This can happen when the uninstall process doesn't fully clean up all registry and cache entries.

To ensure you're running the latest version, manually clear the stale plugin data and reinstall.

Claude Code

1. Edit ~/.claude/settings.json and remove:

   • The "omni-analytics@omni-analytics" entry from enabledPlugins
   • The "omni-analytics" block from extraKnownMarketplaces

2. Edit ~/.claude/plugins/installed_plugins.json and remove the omni-analytics entry

3. Edit ~/.claude/plugins/known_marketplaces.json and remove the omni-analytics entry

4. Delete cached files:

   rm -rf ~/.claude/plugins/cache/omni-analytics/
   rm -rf ~/.claude/plugins/marketplaces/omni-analytics/
   

5. Reinstall the latest version:

   /plugin marketplace add exploreomni/omni-agent-skills
   /plugin install omni-analytics@omni-analytics
   /reload-plugins
   

Cursor

1. Remove the plugin:

   /remove-plugin omni-analytics
   

2. Delete the cached plugin clone:

   rm -rf ~/.cursor/plugins/omni-analytics/
   

3. Reinstall the latest version:

   /add-plugin https://github.com/exploreomni/omni-agent-skills.git
   

Repository Structure

omni-agent-skills/
├── .claude/
│   └── agents/
│       └── sync-cli-changes.md   # maintenance agent, not loaded for plugin users
├── .claude-plugin/
│   ├── marketplace.json
│   └── plugin.json
├── .cursor-plugin/
│   ├── marketplace.json
│   └── plugin.json
├── skills/
│   ├── omni-model-explorer/
│   ├── omni-query/
│   ├── omni-model-builder/
│   ├── omni-content-explorer/
│   ├── omni-content-builder/
│   ├── omni-ai-optimizer/
│   ├── omni-ai-eval/
│   ├── omni-admin/
│   ├── omni-embed/
│   └── omni-integrations/
│       ├── .claude-plugin/
│       │   └── plugin.json
│       ├── .cursor-plugin/
│       │   └── plugin.json
│       └── skills/
│           ├── omni-to-snowflake-semantic-view/
│           ├── omni-to-databricks-metric-views/
│           └── omni-to-dbt-metricflow/
├── agents/
│   ├── omni-analyst.md
│   ├── omni-modeler.md
│   └── omni-admin-agent.md
├── evals/
│   ├── README.md
│   ├── SETUP.md
│   ├── .env.example
│   ├── eval-env.json
│   ├── lib/
│   ├── runner.sh
│   ├── history.sh
│   ├── gepa_export.sh
│   ├── reset.sh
│   └── workspaces/
├── rules/
│   ├── omni-api-conventions.mdc
│   ├── omni-yaml-conventions.mdc
│   └── omni-terminology.mdc
├── assets/
│   ├── logo.svg
│   └── omni-agent-skills-banner.svg
├── CHANGELOG.md
├── README.md
└── LICENSE

Layout Conventions

Skill packages stay compatible with the Agent Skills specification by keeping each skill's SKILL.md at the skill root. Supporting material lives beside it in focused subdirectories such as references/, evals/, scripts/, or assets/; keep references from SKILL.md shallow and relative to the skill root.

The root evals/ directory is contributor tooling, not a distributed skill package. It is a thin BenchFlow wrapper and keeps public entrypoints at the top level:

evals/runner.sh
evals/history.sh
evals/gepa_export.sh
evals/reset.sh

Implementation details are grouped by purpose:

Path	Purpose
evals/lib/	Python helpers used by the BenchFlow runner
evals/.env.example	Template for local model and Omni credentials
evals/eval-env.json	Template for instance-specific eval identifiers
evals/workspaces/	Generated BenchFlow tasks, jobs, trajectories, and summaries
evals/history.sh	CSV/JSONL export of timestamped BenchFlow summaries
evals/gepa_export.sh	GEPA trace export from completed BenchFlow runner workspaces

Per-skill eval definitions live with the skill they evaluate and use BenchFlow's cases schema:

skills/<skill>/evals/evals.json

skills.sh Security Audits

This table is refreshed weekly by .github/workflows/update-skills-security-audits.yml from the public skills.sh audit pages. Status labels use 🟢 Pass, 🟡 Warn, 🔴 Fail, and ⚫ Error.

Skill	Gen Agent Trust Hub	Socket	Snyk
omni-admin	🟢 PassSep 28, 2026, 09:22 AM	🟢 PassSep 28, 2026, 09:23 AM	🟢 PassSep 28, 2026, 09:22 AM
omni-ai-eval	🟢 PassSep 28, 2026, 09:22 AM	🟢 PassSep 28, 2026, 09:22 AM	🟢 PassSep 28, 2026, 09:22 AM
omni-ai-optimizer	🟢 PassSep 28, 2026, 09:22 AM	🟢 PassSep 28, 2026, 09:23 AM	🟢 PassSep 28, 2026, 09:22 AM
omni-content-builder

Documentation

• Skill eval runbook
• Skill eval instance setup
• Omni REST API Reference
• Omni Modeling Documentation
• Optimize Models for Omni AI
• Omni MCP Server
• Claude Code Plugin Docs
• Cursor Plugin Docs
• skills.sh

Contributing

See CONTRIBUTING.md for contributor guidance, including skill and agent authoring, validation, evals, versioning, changelog entries, and PR expectations.

License

Apache 2.0