Skills
Extend your assistant with reusable instructions and multi-step methodologies using the open SKILL.md specification.
A Skill is an isolated folder containing a standardized SKILL.md specification file.
- Templates: Chosen explicitly by users to format a known report or note output;
- Skills: Define trigger descriptions that allow the AI assistant to recognize user intent and dynamically load deep instructions when appropriate.
1. Open the skills directory
In Kapinote, navigate to Settings → Skills and click Open folder. Inside this directory, create one subfolder per skill (naming the folder to match your skill):
skills/
└── customer-follow-up/
└── SKILL.md2. Authoring SKILL.md
The file must begin with YAML frontmatter containing a single-line name and description, followed by detailed Markdown instructions:
---
name: customer-follow-up
description: Use when the user asks for a follow-up email after a customer or sales meeting.
---
# Customer Follow-up Playbook
Follow these core instructions:
1. Use the selected meeting transcript as the sole factual source.
2. Outline key decisions, next milestone dates, and assignees.
3. Do not assume or invent missing commitments.
4. Keep the tone concise, executive, and action-oriented.Design Tip: Use lowercase hyphenated identifiers for
name(e.g.,sprint-retrospective-analysis). Craftdescriptionas an explicit trigger condition explaining when the assistant should load this skill.
3. Reloading skills
After creating or modifying your SKILL.md file, return to Settings → Skills and click Refresh. Valid skills will display their name and description.
If a skill fails to appear:
- Ensure the filename is exactly
SKILL.md(uppercase). - Ensure opening and closing
---fences are valid. - Ensure both
nameanddescriptionare single-line strings. - Ensure the file resides in its own subfolder under
skills/.
4. Skills vs. Templates
| Dimension | Templates | Skills |
|---|---|---|
| Invocation | Manually selected via UI dropdown or / slash command | Dynamically matched and loaded by the assistant based on user intent |
| Storage | Managed in-app and stored in local SQLite | Standard file-based SKILL.md folders |
| Use Case | Fixed output formats (emails, tables, schemas) | Domain playbooks, multi-step reasoning guidelines, and tool workflows |
Best Practice: Keep meeting-specific data out of
SKILL.md. Factual evidence must always originate from the selected meeting transcript.