Back to skills
SKILL.md
003 Name Skill 6310ca0d
ASecurityReference guide for Anthropic official specifications on Skills and Subagents. Use when: (1) Verifying naming conventions, (2) Checking frontmatter field requirements, (3) Validating permission modes, (4) Understanding progressive disclosure patterns. Triggers: "anthropic guide", "official docs", "naming rules", "field requirements"
- 9 stars
- 0 votes
- 0 copies
- 0 views
- Added October 11, 2026
Security analysis
100/100npx -y skills add tools-only/X-Skills --skill 003-name-skill_6310ca0d --agent claude-codeAre you the author of 003 Name Skill 6310ca0d?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tools-only-003-name-skill-6310ca0d)---
name: anthropic-reference
description: |
Reference guide for Anthropic official specifications on Skills and Subagents.
Use when: (1) Verifying naming conventions, (2) Checking frontmatter field requirements,
(3) Validating permission modes, (4) Understanding progressive disclosure patterns.
Triggers: "anthropic guide", "official docs", "naming rules", "field requirements"
---
# Anthropic Official Guide Reference
This Skill provides bundled core content from Anthropic official guidelines.
---
## Skills Guide
→ See [skills-guide.md](references/skills-guide.md) for complete reference
**Core Rules**:
| Field | Constraints |
|-------|-------------|
| `name` | 1-64 chars, lowercase + hyphens, matches directory name |
| `description` | 1-1024 chars, functionality + usage timing |
**Progressive Disclosure**:
1. Metadata (~100 tokens) - Always loaded
2. Instructions (< 5000 tokens) - When activated
3. Resources - When needed
---
## Subagents Guide
→ See [subagents-guide.md](references/subagents-guide.md) for complete reference
**Core Fields**:
| Field | Description |
|-------|-------------|
| `name` | Unique identifier in lowercase + hyphens |
| `description` | Claude auto-delegation trigger |
| `tools` | Allowed tools list |
| `model` | haiku/sonnet/opus/inherit |
| `permissionMode` | Permission mode |
**Permission Modes**:
- `default` - Standard confirmation
- `acceptEdits` - Auto-accept edits
- `plan` - Read-only
- `bypassPermissions` - Skip all confirmations (caution)
---
## Naming Rules
→ See [naming-rules.md](references/naming-rules.md) for validation rules
**Common Rules**:
- 1-64 characters
- Lowercase + numbers + hyphens only
- No hyphen at start/end
- No consecutive hyphens
- No reserved words: `anthropic`, `claude`
**Validation**:
```bash
[[ "$name" =~ ^[a-z][a-z0-9-]*[a-z0-9]$ ]] && [ ${#name} -le 64 ]
```
---
## Quick Reference
### Skill Frontmatter
```yaml
---
name: skill-name
description: |
What it does. Use when: (1) scenario, (2) scenario.
Triggers: "keyword1", "keyword2"
---
```
### Agent Frontmatter
```yaml
---
name: agent-name
description: |
Role description.
Trigger: Use when condition.
tools: [Read, Grep, Glob]
model: sonnet
permissionMode: default
---
```
---
## Validation Scenarios
### name Validation
1. Lowercase + hyphens only?
2. 1-64 characters?
3. No hyphen at start/end?
4. No consecutive hyphens?
5. No reserved words?
6. (Skill) Matches directory name?
### description Validation
1. 1-1024 characters?
2. Includes functionality description?
3. Includes trigger conditions?
4. Includes keywords?
### Tool Permission Validation
1. Least privilege principle?
2. No Write/Edit for Read-only Agents?
3. No Task for Worker Agents?
---
## Bundled References
All guidelines for this Skill are self-contained:
- [skills-guide.md](references/skills-guide.md) - Skills official guide core
- [subagents-guide.md](references/subagents-guide.md) - Subagents official guide core
- [naming-rules.md](references/naming-rules.md) - Naming rules summary
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.