build-moat-assets

name	build-moat-assets
author	JM Labs (Javier Montaño)
description	Create MOAT directory structures and populate asset files (references/, examples/, prompts/, scripts/, assets/) for plugin skills. Trigger: build MOAT assets, create skill assets, populate references, generate examples, build skill directories.
argument-hint	<spec-package-path> <target-plugin-path>
allowed-tools	["Read","Write","Edit","Glob","Grep","Bash"]

Build MOAT Assets

"Structure without content is an empty house; content without structure is a pile of bricks."

Creates MOAT directory structures for plugin skills and populates them with asset files from spec packages. Handles the physical disk operations that turn specification content into a living skill directory with references, examples, prompts, scripts, and static assets.

---

Procedure

Step 1 -- Read MOAT Strategy and Spec Package

• Read the spec package at <spec-package-path>. [Read]
• Extract the list of skills to build, each skill's SKILL.md content, and the MOAT asset file contents (references, examples, prompts, scripts, assets).
• If the spec package is missing SKILL.md content for any skill, halt and report the gap.

Step 2 -- Verify Target Plugin Path

• Confirm <target-plugin-path> exists and contains a skills/ directory. [Glob]
• If the skills/ directory does not exist, create it. [Bash]
• List existing skill directories to detect potential name collisions.

Step 3 -- Create Skill Directories

• For each skill in the spec package, create skills/{skill-name}/. [Bash]
• If the directory already exists, list its contents and warn the user. Do not overwrite without confirmation.
• Use the MOAT directory generation script when available: scripts/generate-moat-dirs.sh. [Bash]

Step 4 -- Write SKILL.md Files

• For each skill, write the SKILL.md content from the spec package into skills/{skill-name}/SKILL.md. [Write]
• Verify the written file is non-empty by checking its existence and line count. [Glob]

Step 5 -- Create MOAT Subdirectories Per Strategy

For each skill, create only the subdirectories that have content:

• references/ -- If reference files exist in the spec.
• examples/ -- If example files exist in the spec.
• prompts/ -- If prompt files exist in the spec.
• scripts/ -- If script files exist in the spec.
• assets/ -- If template/static files exist in the spec.

Do NOT create empty subdirectories. [Bash]

Step 6 -- Write Content Files

For each MOAT subdirectory with content:

• Write each file from the spec package into the appropriate subdirectory. [Write]
• Preserve the file names from the spec package.
• For markdown files: verify they start with an H1 title.
• For JSON files: verify they are valid JSON. [Bash]

Step 7 -- Handle Script Files

For any files in scripts/:

• Verify or add the appropriate shebang line:
  • .sh files: #!/bin/bash
  • .py files: #!/usr/bin/env python3
  • .js files: #!/usr/bin/env node
• Set executable permissions: chmod +x. [Bash]
• Verify the permission was set correctly.

Step 8 -- Write Template and Asset Files

For files in assets/:

• Write template files preserving placeholder markers (e.g., {plugin-name}, {author}).
• For JSON templates: verify valid JSON structure after writing.
• For markdown templates: verify they contain the expected section headers.

Step 9 -- Verify All Files

• Use [Glob] to list all files created under each skill directory.
• Compare the list against the spec package's expected file manifest.
• Report any missing files as CRITICAL.
• Report any extra files as INFO (may be pre-existing).

Step 10 -- Report Creation Summary

Output a summary table:

| Skill | Files Created | Total Lines | MOAT Dirs |
|-------|--------------|-------------|-----------|
| {name} | {count} | {lines} | {dirs} |

Include:

• Total skills processed.
• Total files created.
• Total lines written.
• Any warnings or errors encountered.

---

Quality Criteria

• Every skill directory contains a non-empty SKILL.md file.
• No empty MOAT subdirectories are created.
• All script files have correct shebangs and executable permissions.
• All JSON files are valid JSON.
• The creation summary accounts for every file in the spec package.
• Pre-existing directories are never silently overwritten.

Assumptions & Limits

• Destructive by design (with consent for overwrites). Creates directories and files on disk. Will not overwrite existing skill directories without explicit user confirmation.
• Depends on spec packages being complete. If a spec package is missing MOAT asset content, those subdirectories are skipped (not created empty).
• The generate-moat-dirs.sh script (Step 3) is optional. If not available, directories are created directly via Bash mkdir -p.
• File verification (Step 9) checks existence and non-emptiness only. It does not validate content quality -- that is audit-content-quality's job.
• Windows line ending detection (Edge Case 3) is best-effort. Files with mixed line endings may not be fully corrected.

Good vs Bad

Bad creation summary:

Created 5 skill directories.

Missing: no file counts, no line counts, no MOAT directory details, no warnings.

Good creation summary:

| Skill | Files Created | Total Lines | MOAT Dirs |
|-------|--------------|-------------|-----------|
| validate-hooks | 4 | 210 | references/, examples/, prompts/ |
| validate-manifest | 3 | 145 | references/, examples/ |

Total: 2 skills, 7 files, 355 lines. 0 warnings, 0 errors.

Includes: per-skill breakdown, file counts, line counts, MOAT directories created, totals with error summary.

Anti-Patterns

1. Creating empty directories -- Every MOAT subdirectory must have at least one file. If the spec has no content for a subdirectory, do not create it. Empty directories are noise.
2. Forgetting chmod on scripts -- A script without execute permission will fail silently when invoked. Always set permissions immediately after writing.
3. Writing files without verification -- Disk writes can fail (permissions, disk full, path issues). Always verify each file exists after writing.

Edge Cases

1. Spec package with a single skill and no MOAT assets -- Create just the skill directory with SKILL.md. No subdirectories. Report as valid minimal build.
2. Target path with existing skill of the same name -- List the existing files, warn the user, and wait for confirmation. If confirmed, back up existing SKILL.md as SKILL.md.bak before overwriting.
3. Script files with Windows line endings -- Detect \r\n and convert to \n before writing. Shebangs fail with Windows line endings on Unix systems.