Creating Custom Skills
Skills are domain expertise packages that give agents specialized capabilities — sales methodology, tax optimization, support triage, and more. MeetLoyd follows the Anthropic Agent Skills specification, so skills you build here are portable to any compatible platform.
This guide walks you through creating a custom skill from scratch, structuring it for optimal performance, and publishing it to the Store.
Prerequisites
- A MeetLoyd workspace
- Familiarity with Markdown and YAML frontmatter
Quick Start
1. Create the Skill Directory
Each skill is a directory containing at minimum a SKILL.md file. The directory name must match the name field in your frontmatter.
A typical skill directory looks like this:
| Path | Purpose |
|---|---|
SKILL.md | Required — frontmatter and instructions |
references/ | Optional — detailed docs loaded on demand |
scripts/ | Optional — executable automation (Python, Bash, JS, TS) |
assets/ | Optional — static resources (templates, schemas, images) |
2. Write SKILL.md
Your SKILL.md file has two parts: YAML frontmatter (metadata) and the Markdown body (instructions).
Frontmatter declares the skill's identity, description, and metadata. Body provides the instructions the agent follows when the skill is activated.
A well-structured body includes:
- When to Use This Skill — clear activation triggers and exclusions
- Core Framework — the main methodology, decision tree, or process
- Quick Reference — key facts, tables, or templates needed immediately
- Reference Files — links to detailed docs in the
references/directory
3. Validate Your Skill
MeetLoyd includes a built-in validator that checks your skill against the Anthropic spec. You can run validation from the dashboard under Skills > Validate, or use the reference library from the spec.
The validator checks:
- Name matches directory name
- Description is under 1024 characters
- All referenced files exist
- Frontmatter fields are well-formed
4. Register and Assign
Once your skill directory is in place, the skill registry picks it up automatically on startup. Assign it to an agent from the dashboard:
Agent Settings > Skills > Assign Skill
You can set priority (skills with higher priority are preferred when multiple skills match) and toggle skills on or off without removing the assignment.
Frontmatter Reference
| Field | Required | Constraints |
|---|---|---|
name | Yes | 1-64 chars. Lowercase a-z, 0-9, -. No consecutive hyphens. Must match directory name. |
description | Yes | 1-1024 chars. Describe what the skill does and when to use it. Include discovery keywords. |
license | No | License name or reference to a bundled file. |
compatibility | No | 1-500 chars. Environment requirements. |
metadata.author | No | Organization or individual name. |
metadata.version | No | Semver string (e.g., "1.0.0"). |
metadata.category | No | One of: sales, support, hr, finance, legal, marketing, product, engineering, operations, compliance, general. |
metadata.tags | No | Array of lowercase keywords for search and discovery. |
allowed-tools | No | Space-delimited pre-approved tools (experimental). |
Name Validation Rules
Valid names use lowercase letters, digits, and hyphens. They cannot start or end with a hyphen, and consecutive hyphens are not allowed. Spaces and uppercase letters are rejected.
Progressive Disclosure
Skills use a 3-level loading strategy to minimize context usage:
| Level | What Loads | When | Token Budget |
|---|---|---|---|
| 1. Metadata | Name and description | At startup, for all skills | ~100 tokens |
| 2. Instructions | Full SKILL.md body | When the skill is activated | Less than 5,000 tokens |
| 3. Resources | References, scripts, assets | When specifically needed | As needed |
Keep SKILL.md under 500 lines. Move detailed content (playbooks, battlecards, lookup tables) to reference files. This keeps activation fast and context usage low.
Writing Effective Skills
Strong Trigger Conditions
Tell the agent exactly when to activate the skill and when not to. List specific keywords, phrases, or scenarios that should trigger activation, and explicitly list what should not trigger it.
Decision Trees
Agents work well with structured decision logic. Use clear branching paths (if/then/else) written in plain text. For example, a support triage skill might branch on severity: system down leads to P1, core feature broken leads to P2, workaround available leads to P3 or P4.
Copy-Paste Templates
Provide ready-to-use response templates the agent can adapt. Use placeholders like [NAME], [CONCERN], and [RESPONSE] that the agent fills in contextually.
Concrete Examples
Show input/output pairs — what the customer says, and how the agent should respond. This grounds the skill in real scenarios rather than abstract rules.
Anti-Patterns
Tell the agent what NOT to do. For example: do not discount immediately, do not badmouth competitors, do not overwhelm with features, do not skip the exploration phase.
Real-World Example: MFO Tax Optimization
A production skill from the MFO (Multi-Family Office) vertical demonstrates these best practices:
- Clear activation triggers — activates when the client asks about tax planning, IFI exposure, CDHR impact, life insurance optimization, or asset sale timing
- Structured workflow — 8 sequential steps from fiscal profile gathering through final recommendations
- MCP tools listed — the skill declares which tools the agent can call (income tax calculator, marginal rate lookup, IFI calculator, etc.)
- Concrete parameters — includes specific thresholds and rates so the agent does not have to guess
Testing Skills
Manual Testing
- Assign the skill to a test agent
- Start a conversation that should trigger the skill
- Verify the agent activates the skill (visible in the activity stream)
- Check that reference files load correctly when needed
Validation Checklist
| Check | What to Verify |
|---|---|
| Name match | name field matches directory name |
| Description length | Under 1024 characters with discovery keywords |
| Body length | SKILL.md is under 500 lines |
| File references | All referenced files in references/, scripts/, assets/ exist |
| Markdown links | No broken links |
| Frontmatter | Passes validation with no errors |
| Load test | Skill loads correctly without errors |
Publishing to the Store
Custom skills can be published to the MeetLoyd Store for other tenants:
- Ensure your skill passes all validation
- Go to Store > Publish
- Upload the skill directory
- Set pricing (free or subscription)
- Submit for review
See Store Publishing for details.
Next: Learn about the Skills Architecture to understand how skills are loaded and served internally.