Skill Validator

Skill Validator Skill

Internal skill used by the code-reviewer agent when reviewing skill contributions. Validates a skill file against AI DevStudio standards.

Usage

Called with the path to a skill file to validate. Context will contain the file path.

Validation Checks

1. File Location

Verify the skill is in the correct location:

skills//SKILL.md

Reject if in wrong directory or wrong filename.

2. YAML Frontmatter

Read the skill file and check:

• name field present and matches directory name (kebab-case)
• description field present and non-empty (>10 characters)
• disable-model-invocation field present and is a boolean
• No unknown frontmatter fields (warn on unrecognized keys)

3. Line Count

Bash: wc -l /SKILL.md

Warn if over 100 lines of instruction content (excluding frontmatter, code blocks, Token Optimization section).

Estimate instruction lines:

• Total lines minus frontmatter (lines between ---)
• Minus code block lines (between ``` markers)
• Minus Token Optimization section

4. Token Optimization Section

Check that the file contains a ## Token Optimization section:

Bash: grep -n "## Token Optimization" skills//SKILL.md

Fail if missing. The section must include expected token range.

5. Naming Conventions

• Directory name: kebab-case only
• No underscores, no camelCase
• Name matches name field in frontmatter

6. No Hardcoded Paths

Bash: grep -n "/Users/\|/home/[a-z]\|C:\\Users" skills//SKILL.md

Fail if any absolute user-specific paths found.

Output Format

Skill Validation: skills//SKILL.md

PASS / FAIL / WARNINGS

Checks:
  [PASS] File location
  [PASS] YAML frontmatter
  [WARN] Line count: 112 lines (limit: 100 instruction lines)
  [PASS] Token Optimization section present
  [PASS] Naming conventions
  [PASS] No hardcoded paths

Summary: Ready to merge / Needs fixes

Token Optimization

Expected range: 150-400 tokens

Patterns used: Bash grep for specific checks, early exit on critical failures

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: manastalukdar
• Source: manastalukdar/ai-devstudio
• License: MIT

Install and usage instructions live in the source repository linked above.