
Skill Validator
- 29 installs
- 31 repo stars
- Updated August 2, 2026
- shipshitdev/library
Helps with ai & agent building tasks.
About
skill-validator is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.
- skill-validator
- AI & Agent Building
- AI-coding skill
Skill Validator by the numbers
- 29 all-time installs (skills.sh)
- Ranked #9,369 of 16,546 AI & Agent Building skills by installs in the Skillselion catalog
- Data as of Aug 3, 2026 (Skillselion catalog sync)
npx skills add https://github.com/shipshitdev/library --skill skill-validatorAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 29 |
|---|---|
| repo stars | ★ 31 |
| Last updated | August 2, 2026 |
| Repository | shipshitdev/library ↗ |
What it does
Helps with ai & agent building tasks.
Files
Skill Validator
Validate SKILL.md files against the Agent Skills specification and Claude Code extensions.
When to Run
- After creating a new skill
- After modifying a skill's SKILL.md frontmatter
- Before committing skill changes
- During periodic repo audits
Validation Rules
Required Fields (Agent Skills Spec)
Every SKILL.md must have YAML frontmatter with:
name— kebab-case, matches directory namedescription— 1-3 sentences, under 1024 chars, starts with verb or domain noun
Metadata Block
version and tags must be inside metadata:, never top-level:
# CORRECT
metadata:
version: "1.0.0"
tags: "react, performance, optimization"
# WRONG — top-level version
version: 1.0.0
# WRONG — tags as YAML list
metadata:
tags:
- react
- performanceForbidden Fields
These are not part of any spec:
auto_activate/auto_trigger— removed in 2026-04 migrationrisk— not in Agent Skills or Claude Code specs
Claude Code Extensions (Optional)
Valid extension fields:
| Field | Purpose |
|---|---|
when_to_use | Trigger phrases for auto-activation |
disable-model-invocation | Prevent auto-triggering (for destructive skills) |
user-invocable | Slash command name |
allowed-tools | Pre-approved tool patterns |
model | Model override |
effort | Effort level |
context | fork for subagent isolation |
Content Rules
- No hardcoded
/workspace/paths - No tool names in instructions (say "search for" not "use Grep")
- Imperative/infinitive style ("Configure X" not "You should configure X")
- Code blocks use real backtick fences, not escaped
\\\`
Validation Process
1. Read the SKILL.md frontmatter 2. Check name matches parent directory name 3. Check description exists and is under 1024 chars 4. Check description plus when_to_use is under 1536 chars 5. Check plugin.json description is present and under 100 chars 6. Check version/tags are NOT top-level (must be inside metadata:) 7. Check for forbidden fields (auto_activate, auto_trigger, risk) 8. Check for escaped backtick fences in content 9. Check for hardcoded paths (/workspace/, project-specific paths) 10. Run bunx markdownlint-cli on the file 11. Run ./scripts/validate-skill-sync.sh for cross-validation
Quick Validation Command
# Single skill
bunx markdownlint-cli skills/<name>/SKILL.md skills/<name>/references/*.md
# All skills
bunx markdownlint-cli --ignore bundles --ignore dist "**/*.md"
# Sync validation
./scripts/validate-skill-sync.sh