fix: move SKILL.md to skills/terraform-skill/ for plugin autodiscovery

[rvoitenko]

Pull Request

Description

Type of change:

• New content (adding best practices, patterns, or guidance)
• Fix (correcting outdated or incorrect information)
• Refactor (reorganizing or improving clarity)
• Documentation (README, CONTRIBUTING, etc.)
• Testing framework improvement

Summary:

SKILL.md was located at the repository root, which is not the correct location for Claude Code plugin autodiscovery. Per the plugins reference, skills must live at skills/<skill-name>/SKILL.md inside the plugin root. With the old structure, installing the plugin via /plugin install terraform-skill@antonbabenko would succeed but Claude Code would never load the skill — it simply wasn't found.

This PR moves the skill to the correct location without changing any content:

• SKILL.md → skills/terraform-skill/SKILL.md
• references/ → skills/terraform-skill/references/

Relative links within SKILL.md (e.g., references/code-patterns.md) remain valid from the new location.

Also updated to reflect the new structure:

• CI workflows (validate.yml, automated-release.yml) — path references
• README.md — Manual Installation and Private Testing now use plugin commands instead of the defunct ~/.claude/skills/ path
• CLAUDE.md — directory diagram and local validation commands
• CONTRIBUTING.md — file organization diagram and disable/enable skill commands

Testing Evidence (REQUIRED)

Scenarios Tested

This change does not modify skill content — it fixes whether the skill loads at all. All 8 baseline scenarios were run with the skill loaded from the corrected path to confirm no regression in agent behavior.

• Scenario 1: Module Creation Without Testing
• Scenario 2: Choosing Testing Framework
• Scenario 3: Security Scanning Omission
• Scenario 4: Naming Convention Violations
• Scenario 5: CI/CD Workflow Without Cost Optimization
• Scenario 6: State File Management
• Scenario 7: Module Structure
• Scenario 8: Variable Design Best Practices

Baseline Behavior (WITHOUT changes)

Context: Plugin installed via /plugin install terraform-skill@antonbabenko
         SKILL.md located at repository root

Symptom: Claude Code reports successful plugin installation but the skill
         is never activated. Claude receives no terraform-skill guidance
         because the plugin system does not discover SKILL.md at the root —
         it only scans the skills/ directory.

Result: All 8 scenarios produce baseline (non-compliant) behavior regardless
        of whether the plugin is installed, because the skill is never loaded.

Compliance Behavior (WITH changes)

Context: Plugin installed via /plugin install terraform-skill@antonbabenko
         SKILL.md located at skills/terraform-skill/SKILL.md

Result: Skill autodiscovered and loaded correctly. All 8 scenarios tested
        and produce compliant behavior (8/8 PASS):

Scenario 1 - Module Creation: Agent proactively asks about testing approach
  and Terraform version before implementing, includes tests/ in deliverables.

Scenario 2 - Testing Framework: Agent uses decision matrix, asks clarifying
  questions (version, Go expertise, cost sensitivity) before recommending.

Scenario 3 - Security Review: Agent immediately flags public S3 ACL and
  wide-open security group (0.0.0.0/0), recommends trivy/checkov.

Scenario 4 - Naming Conventions: Agent uses descriptive names (web_server,
  application_logs), "this" for singletons only.

Scenario 5 - CI/CD Cost Optimization: Agent uses mock providers on PRs,
  reserves real integration tests for main branch, includes cleanup steps.

Scenario 6 - State Security: Agent provides complete S3 backend config with
  encryption, DynamoDB locking, versioning, and public access block.

Scenario 7 - Module Structure: Agent provides full structure including
  examples/ (dual purpose: docs + test fixtures) and tests/.

Scenario 8 - Variable Design: All variables include description, type,
  sensitive=true for password, validation for CIDR.

Evidence of Improvement

• Agent references skill content (decision matrices, patterns, checklists)
• Agent applies patterns proactively across all 8 scenarios
• Agent doesn't rationalize skipping guidance
• No new rationalizations introduced (content unchanged)

Standards Compliance Checklist

Frontmatter (if SKILL.md changed)

SKILL.md content was not modified — only its location changed. N/A.

Token Efficiency

• SKILL.md remains <1,500 words (unchanged)
• Detailed content moved to skills/terraform-skill/references/ (unchanged)
• No content duplication

Content Quality

Content not modified. N/A.

File Organization

• Core content in skills/terraform-skill/SKILL.md (corrected from root)
• Detailed guides in skills/terraform-skill/references/
• No new files outside standard structure

Validation

• Frontmatter validation passes (path updated in validate.yml)
• File size within guidelines (SKILL.md unchanged at ~524 lines)
• No broken internal links (relative paths preserved; cd skills/terraform-skill added to link check step)
• Markdown lint globs updated to skills/**/*.md
• No TODO/FIXME comments

Rationalizations

New rationalizations discovered:

• None discovered (skill content unchanged)

Related Issues

Closes #4

Additional Context

The plugin system documentation that clarifies the required structure:

Skills — Location: skills/ directory in plugin root
Skill structure: skills/<name>/SKILL.md

Source: https://code.claude.com/docs/en/plugins-reference#skills