myaidev-method 0.3.4 → 0.3.6

package/USER_GUIDE.md

@@ -28,6 +28,8 @@ This guide covers everything you need to know about using, customizing, and exte
 - [Troubleshooting](#troubleshooting)
 - [Advanced Usage](#advanced-usage)
 - [Best Practices](#best-practices)
+- [Skill Builder (Eval-Driven Development)](#-skill-builder-eval-driven-development)
+- [Contributing Skills to the Marketplace](#-contributing-skills-to-the-marketplace)
 
 ## 🚀 Quick Start
 
@@ -97,7 +99,7 @@ Use the interactive configuration command to set up your environment—no manual
 /configure defaults
 
 # Set up content rules for brand voice
-/myai-content-rules-setup
+/myai-content-rules-config
 ```
 
 ### Immediate Usage
@@ -184,10 +186,11 @@ Skills are organized into packs for modular installation:
 
 | Pack | Skills | Use Case |
 |------|--------|----------|
-| **content** | content-writer, content-verifier, visual-generator, wordpress-publisher | Content creation and publishing |
+| **content** | content-writer, content-verifier, visual-generator, wordpress-publisher, + publishers | Content creation and publishing |
 | **development** | myaidev-workflow, myaidev-architect, myaidev-coder, myaidev-tester, myaidev-reviewer, myaidev-documenter, myaidev-analyze, myaidev-debug, myaidev-refactor, myaidev-performance, myaidev-migrate | SPARC methodology software development |
 | **security** | security-tester, security-auditor | Penetration testing and auditing |
 | **infrastructure** | coolify-deployer, openstack-manager | Deployment and cloud management |
+| **skill-development** | myai-skill-builder, myai-skill-contributor | Skill creation, eval testing, and marketplace submission |
 
 ### SKILL.md Format
 
@@ -219,7 +222,6 @@ Both naming conventions work:
 | `/content-writer` | `/content-writer` | Create content |
 | `/configure` | `/configure` | Configuration |
 | `/wordpress-publisher` | `/wordpress-publisher` | WordPress publishing |
-| `/myai-generate-visual` | `/visual-generator` | Visual generation |
 | `/myai-openstack` | `/openstack-manager` | OpenStack management |
 
 ### Installation Management Commands
@@ -261,7 +263,7 @@ MyAIDev Method uses the `/configure` skill for interactive setup—**no manual `
 | `/configure openstack` | Configure OpenStack VM management |
 | `/configure defaults` | Set default content settings, visual generation API keys |
 | `/configure skills` | List, validate, backup, and manage skills |
-| `/myai-content-rules-setup` | Interactive brand voice and content rules setup |
+| `/content-rules-config` | Interactive brand voice and content rules setup |
 
 ### WordPress Configuration
 
@@ -413,7 +415,7 @@ $ npx myaidev-method@latest update --claude
 📦 Updating components...
 
 Skills:
-  ➕ visual-generator/SKILL.md (new)
+  ➕ myai-visual-generator/SKILL.md (new)
   ✓ content-writer/SKILL.md (unchanged)
 
 📝 wordpress-publisher/SKILL.md has been modified
@@ -518,8 +520,8 @@ Contains all skill definitions as SKILL.md files, organized by category:
 - `skills/docusaurus-publisher/SKILL.md` - Docusaurus documentation
 - `skills/mintlify-publisher/SKILL.md` - Mintlify documentation
 - `skills/astro-publisher/SKILL.md` - Astro site publishing
-- `skills/visual-generator/SKILL.md` - Visual generation
-- `skills/content-coordinator/SKILL.md` - Content coordination
+- `skills/myai-visual-generator/SKILL.md` - Visual generation
+- `skills/content-production-coordinator/SKILL.md` - Content coordination
 
 **Infrastructure:**
 - `skills/coolify-deployer/SKILL.md` - Coolify deployment automation
@@ -620,7 +622,7 @@ Set up your content creation pipeline with these configuration steps:
 
 ```bash
 # Interactive wizard for brand voice, tone, SEO guidelines, and formatting
-/myai-content-rules-setup
+/content-rules-config
 ```
 
 This creates a `content-rules.md` file that customizes all AI-generated content to match your brand. The wizard guides you through:
@@ -665,7 +667,7 @@ Complete workflow from setup to publication:
 
 ```bash
 # 1. First-time setup: Configure brand voice (creates content-rules.md)
-/myai-content-rules-setup
+/content-rules-config
 
 # 2. Configure publishing destination
 /configure wordpress
@@ -694,7 +696,7 @@ For publishing multiple pieces of content at once, use the Content Production Co
 
 ```bash
 # Process all content in a queue directory
-/content-coordinator ./content-queue/
+/content-production-coordinator ./content-queue/
 ```
 
 **Setting Up a Content Queue:**
@@ -752,7 +754,7 @@ This is the section that gets verified for uniqueness.]
 
 3. Run the coordinator:
    ```bash
-   /content-coordinator ./content-queue/
+   /content-production-coordinator ./content-queue/
    ```
 
 #### Coordinator Features
@@ -760,31 +762,31 @@ This is the section that gets verified for uniqueness.]
 **Core Operations:**
 ```bash
 # Verify only, no publishing
-/content-coordinator ./content-queue/ --dry-run
+/content-production-coordinator ./content-queue/ --dry-run
 
 # Skip confirmations (for automation)
-/content-coordinator ./content-queue/ --force
+/content-production-coordinator ./content-queue/ --force
 
 # Detailed progress output
-/content-coordinator ./content-queue/ --verbose
+/content-production-coordinator ./content-queue/ --verbose
 
 # Custom report directory
-/content-coordinator ./content-queue/ --output-dir ./reports/
+/content-production-coordinator ./content-queue/ --output-dir ./reports/
 
 # Higher concurrency for large batches
-/content-coordinator ./content-queue/ --concurrency 5
+/content-production-coordinator ./content-queue/ --concurrency 5
 ```
 
 **Multi-Platform Publishing:**
 ```bash
 # Publish to PayloadCMS instead of WordPress
-/content-coordinator ./content-queue/ --platform payloadcms
+/content-production-coordinator ./content-queue/ --platform payloadcms
 
 # Generate static markdown files
-/content-coordinator ./content-queue/ --platform static --output-dir ./published/
+/content-production-coordinator ./content-queue/ --platform static --output-dir ./published/
 
 # Publish to Docusaurus documentation site
-/content-coordinator ./docs-queue/ --platform docusaurus
+/content-production-coordinator ./docs-queue/ --platform docusaurus
 ```
 
 Each content item can override the default platform using the `target_platform` front matter field.
@@ -792,30 +794,26 @@ Each content item can override the default platform using the `target_platform`
 **Content Rules Integration:**
 ```bash
 # Use custom content rules file
-/content-coordinator ./content-queue/ --content-rules ./brand/content-rules.md
+/content-production-coordinator ./content-queue/ --content-rules ./brand/content-rules.md
 
 # Content rules are auto-detected from standard locations
-/content-coordinator ./content-queue/  # Finds ./content-rules.md automatically
+/content-production-coordinator ./content-queue/  # Finds ./content-rules.md automatically
 ```
 
-The coordinator automatically searches for `content-rules.md` in these locations:
-1. Explicit path via `--content-rules`
-2. `./content-rules.md` (current directory)
-3. `./.claude/content-rules.md`
-4. `./docs/content-rules.md`
+The coordinator reads `content-rules.md` from the project root (alongside `brand-config.json`, `product-config.json`, `service-config.json`, and `content-themes.json`).
 
 **CI/CD Integration with Webhooks:**
 ```bash
 # Send notifications to CI/CD webhook
-/content-coordinator ./content-queue/ --webhook-url https://ci.example.com/webhook
+/content-production-coordinator ./content-queue/ --webhook-url https://ci.example.com/webhook
 
 # Notify on all events including each published item
-/content-coordinator ./content-queue/ \
+/content-production-coordinator ./content-queue/ \
   --webhook-url https://ci.example.com/webhook \
   --webhook-events start,complete,error,published
 
 # Full CI/CD pipeline integration
-/content-coordinator ./content-queue/ --force --webhook-url $CI_WEBHOOK_URL --analytics
+/content-production-coordinator ./content-queue/ --force --webhook-url $CI_WEBHOOK_URL --analytics
 ```
 
 **Available Webhook Events:**
@@ -827,25 +825,25 @@ The coordinator automatically searches for `content-rules.md` in these locations
 **Queue Management:**
 ```bash
 # Add new content to queue without processing
-/content-coordinator ./new-articles/ --add-to-queue
+/content-production-coordinator ./new-articles/ --add-to-queue
 
 # View queue statistics
-/content-coordinator --queue-stats
+/content-production-coordinator --queue-stats
 
 # Process items from queue
-/content-coordinator --queue .content-queue.json
+/content-production-coordinator --queue .content-queue.json
 
 # Clear completed items from queue
-/content-coordinator --clear-completed
+/content-production-coordinator --clear-completed
 ```
 
 **Analytics & Monitoring:**
 ```bash
 # Enable detailed analytics (default)
-/content-coordinator ./content-queue/ --analytics
+/content-production-coordinator ./content-queue/ --analytics
 
 # Full production run with all options
-/content-coordinator ./content-queue/ \
+/content-production-coordinator ./content-queue/ \
   --verbose \
   --output-dir ./reports/ \
   --concurrency 3 \
@@ -867,25 +865,25 @@ When analytics are enabled, generates `analytics-YYYY-MM-DD-HH-MM-SS.md` with:
 
 ```bash
 # Generate crontab entry with instructions
-/content-coordinator --generate-crontab
+/content-production-coordinator --generate-crontab
 
 # Run in cron-safe mode (for crontab)
-/content-coordinator ./content-queue/ --cron --force
+/content-production-coordinator ./content-queue/ --cron --force
 
 # With minimum 6-hour interval between runs
-/content-coordinator ./content-queue/ --cron --min-interval 21600
+/content-production-coordinator ./content-queue/ --cron --min-interval 21600
 ```
 
 **Example crontab entries:**
 ```bash
 # Every 6 hours
-0 */6 * * * cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-coordinator.log 2>&1
+0 */6 * * * cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-production-coordinator.log 2>&1
 
 # Daily at 9 AM
-0 9 * * * cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-coordinator.log 2>&1
+0 9 * * * cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-production-coordinator.log 2>&1
 
 # Weekdays at 9 AM and 5 PM
-0 9,17 * * 1-5 cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-coordinator.log 2>&1
+0 9,17 * * 1-5 cd /path/to/project && npx myaidev-method coordinate-content ./content-queue --cron --force >> /var/log/content-production-coordinator.log 2>&1
 ```
 
 **WordPress Post Scheduling:**
@@ -894,13 +892,13 @@ Schedule posts for future publication using WordPress's native scheduling:
 
 ```bash
 # Delay all posts by 24 hours
-/content-coordinator ./content-queue/ --schedule-delay 24
+/content-production-coordinator ./content-queue/ --schedule-delay 24
 
 # Spread posts 6 hours apart
-/content-coordinator ./content-queue/ --spread-interval 6
+/content-production-coordinator ./content-queue/ --spread-interval 6
 
 # Combined: Start in 24 hours, then every 6 hours
-/content-coordinator ./content-queue/ --schedule-delay 24 --spread-interval 6
+/content-production-coordinator ./content-queue/ --schedule-delay 24 --spread-interval 6
 ```
 
 You can also set specific publish dates in content front matter:
@@ -914,7 +912,7 @@ publish_date: "2026-02-15T09:00:00Z"  # Schedule for specific date
 **Combined Automation (Cron + WordPress Scheduling):**
 ```bash
 # Run via cron, but schedule posts for optimal times
-/content-coordinator ./content-queue/ \
+/content-production-coordinator ./content-queue/ \
   --cron \
   --force \
   --schedule-delay 24 \
@@ -924,7 +922,7 @@ publish_date: "2026-02-15T09:00:00Z"  # Schedule for specific date
 
 #### Checkpoint/Resume
 
-The coordinator automatically saves progress to `.content-coordinator-state.json`. If interrupted:
+The coordinator automatically saves progress to `.content-production-coordinator-state.json`. If interrupted:
 - Re-run the same command to resume from the last checkpoint
 - Use `--fresh` to start over and ignore saved state
 - Delete the state file manually to reset
@@ -1606,10 +1604,9 @@ cat .claude/skills/content-writer/SKILL.md
 
 ```bash
 # Use interactive setup (recommended)
-/myai-content-rules-setup
-
+/content-rules-config
 # Or copy the example template
-cp content-rules.example.md content-rules.md
+cp content-rules-example.md content-rules.md
 ```
 
 **Why use content-rules.md?**
@@ -1953,7 +1950,7 @@ Content rules provide a **non-destructive** way to customize all content generat
 Use the content rules setup command for guided configuration:
 
 ```bash
-/myai-content-rules-setup
+/content-rules-config
 ```
 
 This interactive wizard will guide you through:
@@ -1968,7 +1965,7 @@ This interactive wizard will guide you through:
 
 ```bash
 # Copy the example template
-cp content-rules.example.md content-rules.md
+cp content-rules-example.md content-rules.md
 
 # Edit with your preferences
 nano content-rules.md
@@ -2106,7 +2103,7 @@ When you run any content generation command, the system:
 
 ### Best Practices for Content Rules
 
-1. **Start with the template**: Use `/myai-content-rules-setup` to generate a starting point
+1. **Start with the template**: Use `/myai-content-rules-config` to generate a starting point
 2. **Iterate gradually**: Start with basic rules, refine based on output
 3. **Be specific**: Vague rules produce inconsistent results
 4. **Include examples**: Show what good vs. bad looks like
@@ -2708,218 +2705,273 @@ DEBUG=false
    - Test skill behavior
    - Validate output quality
 
-## 🎁 Contributing Skills to the Marketplace
+## 🔬 Skill Builder (Eval-Driven Development)
 
-Share your custom skills with the community as **Tier 1 open source skills** through the MyAIDev marketplace. All contributed skills go through a stringent review process — automated CI validation checks for security vulnerabilities, destructive commands, and malware patterns, followed by manual maintainer review for quality and usefulness. The contribution system uses a PR-based review workflow.
+The **Skill Builder** (`/myai-skill-builder`) is the recommended tool for creating production-quality skills. It uses eval-driven development — automated testing with parallel A/B comparison, quantitative benchmarking, and iterative refinement — to ensure skills meet marketplace standards before submission.
 
-### Quick Start: Skill Contributor Skill
+### Actions
 
-The easiest way to create and submit skills is using the `/skill-contributor` skill inside Claude Code:
+| Action | Command | Description |
+|--------|---------|-------------|
+| **Create** | `/myai-skill-builder create` | Full guided workflow: concept → SKILL.md → evals → testing → publishing |
+| **Refine** | `/myai-skill-builder refine <path>` | Analyze and improve an existing skill with marketplace quality scoring |
+| **Test** | `/myai-skill-builder test <path>` | Run automated evals with parallel A/B testing |
+| **Benchmark** | `/myai-skill-builder benchmark <path>` | Multi-run statistical analysis with variance measurement |
+| **Validate** | `/myai-skill-builder validate <path>` | Quick structural and security validation |
+| **Publish** | `/myai-skill-builder publish <path>` | Validate, score, and submit to the marketplace |
 
-```bash
-# Interactive menu - choose what you want to do
-/skill-contributor
+### Create Workflow (11 Phases)
 
-# Create a new skill with guided prompts
-/skill-contributor create
+The `create` action walks through a complete skill development lifecycle:
 
-# Validate and submit an existing skill via PR
-/skill-contributor submit
+1. **Concept Discovery** — Interactive questions about purpose, target user, and tools needed
+2. **Identity** — Name, description (with "when" clause), and category
+3. **Generate SKILL.md** — Creates from a template tier (basic / intermediate / advanced with sub-agents)
+4. **Progressive Disclosure** — Moves long reference docs to `references/`, deterministic tasks to `scripts/`, templates to `assets/`
+5. **Iterative Refinement** — Self-review against quality checklist, user feedback loop
+6. **Validate** — Runs `npx myaidev-method addon validate` to check structure and security
+7. **Write Evals** — Generates automated test cases in `evals/evals.json` with assertions
+8. **Run & Grade** — Parallel A/B testing (with-skill vs baseline) with grader subagent scoring
+9. **Iterate** — Improve skill based on eval results, rerun to verify
+10. **Description Optimization** — Train/test split on trigger queries, optimize for activation accuracy
+11. **Publish Decision** — Submit to marketplace with eval evidence
 
-# Check the status of your submissions
-/skill-contributor status
-```
+### Eval-Driven Testing
 
-### Complete Contribution Workflow
+The skill builder runs automated evaluations to measure whether a skill actually helps:
 
-#### Step 1: Authenticate
+```
+{slug}-workspace/
+├── iteration-1/
+│   ├── eval-basic-create/
+│   │   ├── with_skill/           # Outputs from skill-assisted run
+│   │   │   ├── outputs/
+│   │   │   ├── timing.json       # {total_tokens, duration_ms}
+│   │   │   └── grading.json      # Per-assertion PASS/FAIL with evidence
+│   │   └── without_skill/        # Outputs from vanilla Claude baseline
+│   │       ├── outputs/
+│   │       ├── timing.json
+│   │       └── grading.json
+│   └── benchmark.json            # Aggregated pass rates, token usage, timing
+├── iteration-2/
+│   └── ...
+├── benchmark.json                # Multi-run statistical results
+└── feedback.json                 # User feedback from HTML reviewer
+```
 
-Before submitting skills, authenticate with the marketplace:
+For each eval, **two parallel agents** are spawned: one with the skill loaded and one without (baseline). A **grader subagent** then evaluates each output against assertions and provides PASS/FAIL verdicts with cited evidence.
+
+The `benchmark` action runs 3+ iterations to measure variance:
 
 ```bash
-npx myaidev-method login
+/myai-skill-builder benchmark .claude/skills/my-skill
 ```
 
-This opens your browser for OAuth authentication and stores a token locally.
+This produces statistical analysis (mean, standard deviation) for pass rates, token usage, and timing, plus identifies flaky evals and non-discriminating assertions via an **analyzer subagent**.
 
-#### Step 2: Create Your Skill
+### Supporting Files
 
-**Option A: Using the skill-contributor skill (Recommended)**
+Skills can include supporting directories that travel with the skill through the entire pipeline (submit → store → download → install):
 
-```bash
-/skill-contributor create
-```
+| Directory | Purpose | Examples |
+|-----------|---------|---------|
+| `agents/` | Sub-agent definitions for Task tool delegation | `grader-agent.md`, `analyzer-agent.md` |
+| `scripts/` | Deterministic computation, validation, aggregation | `validate.py`, `aggregate_benchmark.py` |
+| `references/` | API schemas, long guides, documentation | `schemas.md`, `api-spec.md` |
+| `assets/` | Templates, HTML files, starter configs | `eval_review.html`, `template.json` |
+| `evals/` | Automated test definitions | `evals.json` |
 
-You'll be prompted for:
-- **Name**: lowercase with hyphens (e.g., `code-formatter`)
-- **Description**: Include a "when to use" clause
-- **Category**: development, content, publishing, security, deployment, infrastructure, or other
-- **Complexity**: basic, intermediate, or advanced (determines template)
+When the skill builder detects that agents repeatedly write the same helper code during evals, it extracts the pattern into a bundled script in `scripts/`.
 
-**Option B: Manual creation**
+### Marketplace Quality Scoring
 
-Create a directory and SKILL.md file:
+Skills are scored on 5 categories (0-100 each) both by the skill builder (locally) and by the marketplace's AI analysis (server-side):
+
+| Category | Weight | What It Measures |
+|----------|--------|-----------------|
+| **Quality** | 25% | Instruction clarity, structure, formatting, agent persona specificity |
+| **Security** | 20% | No credentials, no destructive commands, safe patterns |
+| **Completeness** | 25% | All sections filled, error handling, edge cases, Quick Start |
+| **Originality** | 15% | Distinct problem, genuine value, not a trivial wrapper |
+| **Usefulness** | 15% | Real user need, well-scoped, time savings vs manual approach |
+
+Preview your score before submitting:
 
 ```bash
-mkdir -p .claude/skills/my-skill
+/myai-skill-builder refine .claude/skills/my-skill
 ```
 
-Create `.claude/skills/my-skill/SKILL.md`:
+### Skill Writing Best Practices
+
+- **Progressive disclosure**: Keep SKILL.md under 500 lines. Move reference material to `references/`, deterministic tasks to `scripts/`.
+- **Explain the why**: Don't just say "do X" — explain why X matters so the agent can adapt when circumstances change.
+- **Avoid heavy-handed MUSTs**: Describe goals and constraints. The agent is intelligent — let it figure out how.
+- **Include "when" contexts in description**: "Use when building X, debugging Y, or preparing Z." Undertriggering is worse than overtriggering.
+- **Keep it lean**: After each iteration, remove instructions that don't contribute to passing evals.
 
-```markdown
----
-name: my-skill
-description: "Does X when you need to Y. Use when working with Z."
-argument-hint: "[action] [args]"
-allowed-tools: [Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion]
-context: fork
 ---
 
-# My Skill
+## 🎁 Contributing Skills to the Marketplace
 
-You are a **My Skill** agent — description of what you do.
+The MyAIDev Marketplace at `dev.myai1.ai` hosts community and first-party skills. All contributed skills go through automated AI analysis and admin review before being published.
 
-## Quick Start
+### Submission Pipeline
 
 ```
-/my-skill action args
+Author creates skill
+  → CLI validates locally (structure, security, frontmatter)
+  → CLI bundles SKILL.md + supporting files
+  → API validates bundle (path traversal, size limits, content scanning)
+  → AI analysis scores the skill (quality, security, completeness, originality, usefulness)
+  → Admin reviews and approves/rejects
+  → Published and installable via `addon install`
 ```
 
-## Arguments
+### Quick Start
 
-Parse action from: `$ARGUMENTS`
+**Option A: Using the Skill Builder (Recommended)**
 
-### Actions
-
-#### `action1` - Primary Action
+```bash
+# Full guided workflow with eval testing
+/myai-skill-builder create
 
-**Workflow:**
-1. Step one
-2. Step two
-3. Step three
+# Or submit an existing skill
+/myai-skill-builder publish .claude/skills/my-skill
+```
 
-## Error Handling
+**Option B: Using the Skill Contributor**
 
-- Explain errors and suggest alternatives
-- Never silently skip steps
+```bash
+# Quick scaffolding and submission
+/skill-contributor create
+/skill-contributor submit
 ```
 
-#### Step 3: Validate Your Skill
-
-Run validation to check for issues before submitting:
+**Option C: CLI directly**
 
 ```bash
-# Via CLI
+npx myaidev-method login
 npx myaidev-method addon validate --dir .claude/skills/my-skill
+npx myaidev-method addon submit --dir .claude/skills/my-skill
+npx myaidev-method addon status
+```
 
-# Via skill-contributor
-/skill-contributor submit
-# (validates automatically before submission)
+### What Gets Submitted
+
+When you submit, the CLI automatically detects and bundles supporting files alongside the SKILL.md:
+
+```
+my-skill/
+├── SKILL.md                       ← submitted as content
+├── agents/grader-agent.md         ← bundled as supportingFiles
+├── scripts/validate.py            ← bundled as supportingFiles
+├── references/schemas.md          ← bundled as supportingFiles
+└── evals/evals.json               ← bundled as supportingFiles
 ```
 
-**Validation checks:**
-- ✅ Required frontmatter fields (`name`, `description`, `allowed-tools`)
-- ✅ Content structure (H1 heading, sections)
-- ✅ Security patterns (no credential paths, no destructive commands)
-- ✅ File size limits (50KB max per file, 200KB max total)
-- ✅ Name availability (not already taken in marketplace)
+The CLI skips `__pycache__`, `.pyc`, `.DS_Store`, and `node_modules` when collecting files.
 
-#### Step 4: Submit for Review
+### Bundle Validation (Server-Side)
 
-```bash
-# Via CLI
-npx myaidev-method addon submit --dir .claude/skills/my-skill
+The server validates all submitted bundles:
 
-# Via skill-contributor
-/skill-contributor submit
-```
+- **Path traversal prevention** — No `../` or absolute paths
+- **Allowed directories only** — Files must be in `agents/`, `scripts/`, `references/`, `assets/`, or `evals/`
+- **Blocked file types** — No `.exe`, `.dll`, `.sh`, `.bat`, `.env`, `.pem`, `.key`
+- **Size limits** — Max 500KB total for supporting files, max 50 files
+- **Content scanning** — All files scanned for dangerous patterns (credential references, destructive commands, remote code execution)
+
+### Submission Statuses
 
-This will:
-1. Fork the [myaidev-marketplace](https://github.com/myaione/myaidev-marketplace) repo
-2. Create a branch with your skill
-3. Open a PR for review
-4. Record your submission in the system
+| Status | Meaning |
+|--------|---------|
+| `pending_review` | Submitted, waiting for processing |
+| `ai_analyzing` | AI is scoring the skill automatically |
+| `ready_for_review` | AI analysis complete, waiting for admin review |
+| `approved` | Published to the marketplace |
+| `rejected` | Did not meet quality standards |
+| `changes_requested` | Admin requested modifications |
 
-#### Step 5: Track Your Submission
+Track your submissions:
 
 ```bash
-# Check all your submissions
+# All submissions
 npx myaidev-method addon status
 
-# Or via skill-contributor
-/skill-contributor status
+# Specific submission by ID
+npx myaidev-method addon status <submission-id>
 ```
 
-### Skill Quality Guidelines
+### Installing Bundled Skills
 
-#### Frontmatter Requirements
+When a skill has supporting files, the CLI downloads everything as a JSON bundle and recreates the full directory structure:
 
-| Field | Required | Description |
-|-------|----------|-------------|
-| `name` | ✅ | Lowercase, hyphens only, max 64 chars |
-| `description` | ✅ | Must include "when" clause explaining usage |
-| `allowed-tools` | ✅ | Array of tools the skill needs |
-| `argument-hint` | Recommended | Shows usage pattern in help |
-| `context` | Recommended | Usually `fork` for isolation |
+```bash
+npx myaidev-method addon install my-skill
+```
 
-#### Content Structure
+Creates:
 
-- **H1 heading**: Must match skill purpose
-- **Quick Start**: Show basic usage example
-- **Arguments section**: Document all arguments and actions
-- **Workflow steps**: Clear numbered steps for each action
-- **Error handling**: Guidance for edge cases
+```
+.claude/skills/my-skill/
+├── SKILL.md
+├── agents/
+│   └── grader-agent.md
+├── scripts/
+│   └── validate.py
+└── references/
+    └── schemas.md
+```
+
+Skills without supporting files are downloaded as plain markdown for backward compatibility.
 
-#### Security Rules
+### Skill Requirements
 
-Your skill must NOT contain:
-- ❌ Credential paths (`~/.ssh`, `~/.aws`, etc.)
-- ❌ Destructive commands (`rm -rf /`, `chmod 777`, etc.)
-- ❌ Dynamic code execution (`eval()`, `exec()`) unless essential
-- ❌ Remote script piping (`curl | bash`)
+#### Frontmatter
 
-### Skill Templates
+| Field | Required | Rules |
+|-------|----------|-------|
+| `name` | Yes | Lowercase with hyphens, max 64 chars |
+| `description` | Yes | Must include "when" clause (e.g., "Use when...") |
+| `allowed-tools` | Yes | Only tools the skill actually uses |
+| `argument-hint` | Recommended | Shows expected argument format |
+| `context` | Recommended | Use `fork` for most skills |
 
-The skill-contributor provides three templates based on complexity:
+#### Content Structure
 
-| Template | Best For | Features |
-|----------|----------|----------|
-| **Basic** | Simple, single-action skills | Minimal structure, quick to fill |
-| **Intermediate** | Multi-action skills | Multiple actions, quality checks |
-| **Advanced** | Orchestrator skills | Sub-agents via Task tool, configuration |
+- H1 heading matching skill purpose
+- Quick Start section with concrete usage example
+- Arguments section documenting `$ARGUMENTS` parsing
+- Workflow with numbered steps for each action
+- Error handling with realistic scenarios
+- SKILL.md body under 500 lines
 
-### After Submission
+#### Security Rules (violations block submission)
 
-1. **CI Validation**: GitHub Actions validates your SKILL.md automatically
-2. **Review**: Maintainers review for quality, security, and usefulness
-3. **Feedback**: You may receive comments requesting changes
-4. **Merge**: Once approved, your skill is merged and ingested into the marketplace
-5. **Published**: Users can install via `npx myaidev-method addon install your-skill`
+- No credential paths (`~/.ssh`, `~/.aws`, `/etc/shadow`)
+- No destructive commands (`rm -rf /`, `chmod 777`, `mkfs`)
+- No remote script piping (`curl | bash`, `wget | sh`)
+- Minimal dynamic code evaluation (`eval()`, `exec()`)
 
-### Example: Creating a Simple Skill
+### Example: End-to-End Skill Creation
 
 ```bash
-# 1. Login
+# 1. Authenticate
 npx myaidev-method login
 
-# 2. Create skill interactively
-/skill-contributor create
-# Enter: name=code-formatter, description="Formats code files. Use when cleaning up code style.", category=development, complexity=basic
-
-# 3. Edit the generated SKILL.md with your logic
-code .claude/skills/code-formatter/SKILL.md
-
-# 4. Test locally by invoking it
-/code-formatter "src/**/*.js"
+# 2. Build skill with eval-driven testing
+/myai-skill-builder create
+# → Concept discovery questions
+# → Generates SKILL.md + supporting files
+# → Writes evals, runs A/B tests
+# → Optimizes description for trigger accuracy
+# → Submits to marketplace with eval evidence
 
-# 5. Validate
-npx myaidev-method addon validate --dir .claude/skills/code-formatter
-
-# 6. Submit
-/skill-contributor submit
+# 3. Check status
+npx myaidev-method addon status
 
-# 7. Check status
-/skill-contributor status
+# 4. After approval, others can install
+npx myaidev-method addon install your-skill-name
 ```
 
 ## 📚 Additional Resources
@@ -2950,4 +3002,4 @@ If you encounter issues:
 
 **Happy customizing!** 🚀
 
-Remember: The power of MyAIDev Method lies in its customizability. Don't hesitate to experiment with different skill configurations to find what works best for your workflow.
+Remember: The power of MyAIDev Method lies in its customizability. Don't hesitate to experiment with different skill configurations to find what works best for your workflow.