validate-structure

name	validate-structure
description	Validates plugin directory layout against the official Claude Code plugin spec. Triggers: validate structure, check layout, directory audit, plugin skeleton check.
author	JM Labs (Javier Montaño)
allowed-tools	["Read","Glob","Grep","Bash"]

Validate Structure

"A plugin that does not follow the spec is not a plugin -- it is a folder with aspirations."

Validates that a Claude Code plugin's directory tree conforms to the official specification. Detects missing directories, misplaced files, naming violations, path traversal risks, and script permission issues.

---

Procedure

Step 1 -- Locate Plugin Root

• Search for plugin.json at the provided path (official spec location).
• If not found at root, check .claude-plugin/plugin.json (legacy location). Report INFO: plugin.json found at legacy .claude-plugin/ path if found there.
• If not found at either location, report CRITICAL: No plugin.json found -- not a valid plugin root.
• Record the plugin root path for all subsequent steps.

Step 2 -- Validate Plugin Root Contents

• If .claude-plugin/ directory exists, list all files inside it.
• If plugin.json is inside .claude-plugin/, only plugin.json should be there. Any other file is a WARNING: Unexpected file in .claude-plugin/.
• If plugin.json is at root (preferred), this check is N/A.

Step 3 -- Check Required Directories

• Verify existence of: skills/, commands/, agents/, hooks/.
• Per the official spec, all directories are optional. A plugin needs only plugin.json.
• Missing skills/ = WARNING if the plugin description implies skill-based functionality.
• Missing commands/ = INFO (commands are optional).
• Missing agents/ = INFO (plugin can use the host agent).
• Missing hooks/ = INFO (hooks are optional).

Step 4 -- Validate Component Layout

• Skills MUST be directories containing a SKILL.md file: skills/<name>/SKILL.md.
• Commands MUST be .md files directly under commands/: commands/<name>.md.
• Agents MUST be .md files directly under agents/: agents/<name>.md.
• Report CRITICAL for any skill directory missing SKILL.md.
• Report WARNING for unexpected file types (.txt, .json) in commands/ or agents/.

Step 5 -- Validate Naming Conventions

• All directory and file names (excluding extensions) MUST be kebab-case: ^[a-z][a-z0-9]*(-[a-z0-9]+)*$.
• Report WARNING for each violation with the offending name and suggested kebab-case alternative.

Step 6 -- Scan for Path Traversal and Absolute Paths

• Grep all .md, .json, .sh files for ../ sequences.
• Grep for absolute paths matching ^/ in script files and hook configurations.
• Report CRITICAL for any ../ usage in hooks.json or script references.
• Report WARNING for ../ in documentation files (may be intentional examples).

Step 7 -- Check Script Files

• Locate all .sh, .py, .js files in the plugin tree.
• Verify each has an appropriate shebang line (#!/bin/bash, #!/usr/bin/env node, etc.).
• Verify executable permission is set (chmod +x). Use ls -la to check.
• Report WARNING for missing shebang. Report WARNING for missing execute permission.

Step 8 -- Generate Structure Report

• Produce a findings table: | Severity | File/Path | Finding | Spec Reference |.
• Include summary counts: total CRITICAL, WARNING, INFO.
• If zero CRITICAL findings, state: "Structure validation PASSED".

---

Quality Criteria

• Every finding references the exact path relative to plugin root.
• Severity is never inflated -- follow the classification matrix from the agent.
• Naming suggestions are provided for every kebab-case violation.
• The report is machine-parseable (table format, consistent severity labels).
• plugin.json is searched at root first, then .claude-plugin/ (official spec order).

Assumptions & Limits

• Read-only. This skill never creates directories or modifies files.
• Cannot determine whether a missing directory is intentional (minimal plugin) or accidental (incomplete scaffold). Reports severity based on functionality implications, not intent.
• Symlinks are detected but not followed -- the skill cannot validate the symlink target.
• Does not validate file contents (that is validate-manifest's and validate-components' job). Only checks existence, location, and naming.
• Path traversal detection uses pattern matching, not symlink resolution -- a ../ that resolves safely is still flagged because it indicates a portability risk.

Good vs Bad

Bad finding:

Missing skills directory.

Missing: no severity, no path, no spec reference.

Good finding:

WARNING | skills/ | Directory absent. Plugin description implies skill-based functionality but no skills/ directory exists. | Official spec: optional

Includes: severity, path, context explaining why it matters, spec reference.

Anti-Patterns

1. Reporting a missing optional directory as CRITICAL.
2. Failing to check inside skill subdirectories for SKILL.md.
3. Ignoring dotfiles (.gitkeep, .DS_Store) -- these should be INFO, not errors.
4. Checking naming conventions on files outside the plugin tree.

Edge Cases

1. Plugin with zero skills but valid structure (commands-only plugin) -- valid, INFO note.
2. Nested skill directories (skills/group/skill-name/) -- not spec-compliant, CRITICAL.
3. Symlinks in the plugin tree -- report as WARNING, do not follow.
4. Empty directories (created but no content) -- WARNING for skills/, INFO for others.
5. Mixed case in .claude-plugin directory name -- CRITICAL on case-sensitive filesystems.