メニュー
A personal sandbox skill used as the annotated authoring template for this reference implementation. It shows the agentskills.io-recommended SKILL.md structure. Use it when learning how to author a new skill.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/databricks-solutions/ai-governance --skill hello-skillsコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
| name | hello-skills |
| description | A personal sandbox skill used as the annotated authoring template for this reference implementation. It shows the agentskills.io-recommended SKILL.md structure. Use it when learning how to author a new skill. |
Annotated template (Tier 1 sandbox). This file is the canonical starting point for a new skill. Each section below names an agentskills.io best-practice category and explains what to put there. Replace every
TODO:with real content, then promote viascripts/new-skill.sh/ a PR. Tier-1 skills are never synced and are exempt from the no-TODOgate — a promoted (T2/T3) skill must have allTODOs filled.
TODO: One paragraph — the single task this skill handles and WHEN the agent should reach for it.
Keep scope narrow (one skill, one job); overlapping skills cause mis-routing. Frontmatter
description = what it does + when to use it + trigger keywords (this drives auto-load).
TODO: Real phrasings a user might say — these calibrate description-based auto-load:
TODO: Explicit, ordered steps. Calibrate control to fragility: be prescriptive for fragile or must-be-exact sequences; give the agent freedom (explain why, not rigid steps) where multiple approaches are valid. Provide a default, not a menu of equal options.
TODO: Concrete input → output. Agents pattern-match against examples better than prose. Request: "TODO: a sample question a user would ask" Expected behavior: TODO: what the skill does — the SQL it runs and/or the shape of the answer.
TODO: Common variations/exceptions and how to handle them.
TODO: Environment-specific facts the agent would get WRONG without being told — not general advice. Highest-value content in many skills.
TODO: The skill's data footprint, in prose. Runtime access is governed by Unity Catalog on whoever runs the skill — this documents intent, it is NOT an access control.
catalog.schema.table — what it is read forAnalyze Enterprise Analytics & BI adoption from system.query.history — active users, query volumes, warehouse utilization, and slow-query hotspots. Ask me which dashboards/warehouses are busiest or where query performance is degrading.
Track Advanced Analytics model performance and inference volumes from ML monitoring tables — accuracy/drift metrics by model and version, plus serving request volumes. Ask me which models are degrading or which endpoints are busiest.
Report Unity Catalog governance coverage — table comments, ownership, tags, and column documentation — from system.information_schema. Ask me which tables lack owners, descriptions, or classification tags across a catalog.