[Announcement]: 🤖 New: AI agent skill for the te CLI (Claude, Copilot, and others)

[gronnerup]

Announcement type

Feature

CLI version (if applicable)

No response

Summary

Drop-in AI agent skill that teaches Claude, GitHub Copilot, and other AI coding agents how to use the Tabular Editor CLI. Once installed, your agent writes idiomatic te commands instead of guessing or hallucinating flags. Ships as a single Markdown file with per-agent install instructions.

Details

What's a "skill"?

A skill is a single Markdown file an AI agent loads on demand. It teaches the agent the conventions, command reference, common workflows, and gotchas for a specific tool. Once installed, you can ask things like "deploy this model to my Fabric workspace" or "add a margin measure with a percent format" and the agent reaches for the right te commands by itself, instead of guessing flag names or making up syntax.

What this skill covers

• All te commands across all families: Model I/O, editing, inspection, analysis & quality, execution, deployment & refresh, testing, connection & auth, configuration, shell.
• Authentication patterns: interactive browser, service principal (secret + certificate), env-var, managed identity.
• Object path grammar: slash form (Sales/Revenue), DAX form ('Sales'[Revenue]), wildcards for filter paths.
• Staging model (--save / --stage / --revert) and the interactiveEditMode config.
• TE2 to new CLI migration mapping and the TE_COMPAT=te2 compat layer.
• CI/CD recipes for GitHub Actions and Azure DevOps Pipelines.
• The gotchas that trip up agents in real-world use (partition path asymmetries, MExpression vs expression, BPA JSON output shape, --output-format vs --serialization, and more).

Where it works

Step-by-step installation is documented for:

• Claude Code (project and user scope)
• Claude.ai / Claude Desktop (via the built-in Skills UI)
• GitHub Copilot (repository and VS Code user scope)
• Generic agents using the AGENTS.md convention (Aider, Continue, OpenAI Codex CLI, custom in-house agents)

Each agent has its own loading semantics and trade-offs. For example, Claude loads the skill on demand only when the prompt matches; Copilot injects it into every interaction. The skill README walks through the practical implications for each.

Try it

1. Install the skill for your agent of choice (about a minute, mostly clicks).
2. Ask the agent to do something semantic-model-flavored: "deploy ./model to MyWorkspace", "add a Total Revenue measure", "run BPA and tell me which rules my model fails".
3. Watch it write the right te commands.

Preview

The Tabular Editor CLI is in Limited Public Preview, and so is this skill. As the CLI surface evolves between now and GA, the skill will keep pace. See the CHANGELOG (linked below) for what's in this initial release.

Feedback

Bugs, gaps, missing patterns, wrong examples are all welcome. Open an issue and apply the skill label so we can triage them separately from CLI behavior reports.

Related links

Skill folder: https://github.com/TabularEditor/CLI/tree/main/skill
Per-agent install guide: https://github.com/TabularEditor/CLI/blob/main/skill/README.md
Changelog: https://github.com/TabularEditor/CLI/blob/main/skill/CHANGELOG.md
Release notes: https://github.com/TabularEditor/CLI/releases/tag/skill-v0.1.0
CLI announcement blog post: https://tabulareditor.com/blog/introducing-the-tabular-editor-cli-limited-public-preview
Anthropic Skills documentation: https://docs.claude.com/en/docs/claude-code/skills