| 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 via
scripts/new-skill.sh / a PR. Tier-1 skills are never synced and are exempt from the no-TODO
gate — a promoted (T2/T3) skill must have all TODOs filled.
hello-skills
Overview
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).
When to use this skill
TODO: Real phrasings a user might say — these calibrate description-based auto-load:
- "TODO: an example request that should load this skill"
Instructions
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: first step
- TODO: next step
Examples
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.
Edge cases
TODO: Common variations/exceptions and how to handle them.
Gotchas
TODO: Environment-specific facts the agent would get WRONG without being told — not general
advice. Highest-value content in many skills.
- TODO: e.g. "table X uses soft deletes; filter WHERE deleted_at IS NULL"
Data scope
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 for- No PII or regulated data is accessed.
