| name | office-skill-authoring |
| description | Use when creating or reviewing Office AI Skills. Enforces SKILL.md directory structure, concise YAML metadata, progressive disclosure, references, scripts, app scope, and allowed tool boundaries. |
| application | Excel,Word,PowerPoint |
| tags | skill, skills, authoring, SKILL.md, progressive-disclosure, harness, agent-loop |
| allowed-tools | memory.search, memory.list_recent |
| intent_types | skill_authoring, architecture |
Office Skill Authoring
Use this skill when adding or reviewing a Skill for Office AI Agent.
Required Structure
Each directory-based Skill must use:
skill-name/
├── SKILL.md
├── references/ optional detailed docs, loaded only after the skill is selected
├── scripts/ optional executable helpers
└── assets/ optional templates or examples
SKILL.md must start with YAML front matter. Required fields:
---
name: skill-name
description: One concise sentence that explains when this skill should be used.
---
Recommended fields for this project:
application: Excel
tags: excel, formula, chart
allowed-tools: ApplyFormula, CreateChart
intent_types: data_analysis, formula
Rules
- Keep the description action-oriented; it is used for first-pass skill selection.
- Put long examples and domain rules in
references/ so the first prompt remains small.
- Restrict
allowed-tools to the smallest useful set.
- State app scope clearly: Excel, Word, PowerPoint, or common.
- A Skill should teach the agent how to decide and execute; it should not be a pile of keyword routes.
- Scripts must live in
scripts/ and declare arguments in nearby docs when needed.