What Is the AI Shipping Kit? Purpose and Architecture in the PM Skills Ecosystem

The AI Shipping Kit is a plugin-style module in the phuryn/pm-skills repository that equips product managers and founders with a repeatable, documented workflow for converting AI-generated applications into audited, test-validated shipping packets ready for human review.

The AI Shipping Kit (pm-ai-shipping) closes the critical gap between rapid AI-assisted development and production safety. As part of the broader PM Skills ecosystem maintained in phuryn/pm-skills, this plugin addresses the "last mile" problem of vibe-coded applications by enforcing documentation, security audit, and test-coverage standards before release.

Architectural Overview of the AI Shipping Kit

The kit follows a plugin-based architecture that registers commands under the /pm-ai-shipping:<command> namespace. This design allows seamless integration with AI coding assistants while maintaining strict separation between documentation generation, validation logic, and final packet assembly.

Plugin Entry Point and Metadata

The AI Shipping Kit exposes itself as a Claude plugin defined in pm-ai-shipping/.claude-plugin/plugin.json. This metadata file registers the plugin's capabilities and enables the slash-command namespace that users interact with when orchestrating shipping workflows. The plugin occupies a distinct directory (pm-ai-shipping/) within the repository, isolating its skills and commands from other PM Skills modules.

Core Skills: Documentation and Validation

Two specialist skills provide the intellectual foundation for the kit's operations:

• shipping-artifacts – Defines the mandatory and conditional documentation set required for safe deployment. This includes core artifacts (architecture diagrams, data flows, permission matrices, environment variables, and test specifications) plus conditional docs for email integrations, cron jobs, SEO metadata, and automation workflows. The skill specification lives in skills/shipping-artifacts/SKILL.md.

• intended-vs-implemented – Evaluates the gap between documented product intent and actual code implementation, flagging deviations that could introduce regression risks or security holes. This validation logic is documented in skills/intended-vs-implemented/SKILL.md.

Command Orchestration Layer

Five discrete commands orchestrate the end-to-end shipping process, each defined as a separate Markdown file in the commands/ directory:

1. /ship-check – Runs the complete validation sequence and assembles the final shipping packet (defined in commands/ship-check.md).
2. /document-app – Reverse-engineers the existing codebase into the standardized documentation set.
3. /derive-tests – Maps existing test coverage against documented business rules and identifies validation gaps.
4. /security-audit-static – Performs static security analysis cross-referenced against the permissions and architecture documentation.
5. /performance-audit-static – Identifies N+1 queries, over-fetching patterns, and missing database indexes.

The Shipping Packet Output

The final artifact generated by /ship-check is a shipping packet—a comprehensive Markdown document containing:

• Documentation inventory tracking presence or absence of required artifacts
• Agent context files (CLAUDE.md/AGENTS.md) for future AI development cycles
• Test-coverage mapping with identified gaps against business requirements
• Security and performance audit summaries with severity classifications
• Launch blockers and prioritized next-action recommendations

This packet serves as the single source of truth for human sign-off and provides structured context to feed back into subsequent AI-driven development iterations.

Running the AI Shipping Kit: Command Reference

Users interact with the kit through slash commands that can be executed individually or as a coordinated sequence. Below are the primary invocation patterns as implemented in the source code:

# Generate core documentation from existing code

/pm-ai-shipping:document-app

# Run static security analysis

/pm-ai-shipping:security-audit-static

# Identify performance anti-patterns

/pm-ai-shipping:performance-audit-static

# Map test coverage against requirements

/pm-ai-shipping:derive-tests

To execute the complete workflow in a single step:

# Assemble full shipping packet with all validations

/pm-ai-shipping:ship-check

The resulting packet follows this standardized structure:

## Shipping Packet: my-awesome-app

### Documentation Inventory

| Doc | Status | Notes |
|-----|--------|-------|
| architecture.md | present | – |
| flows.md | present | – |
| permissions.md | missing | Required for auth features |

### Security Summary

- Critical: 2 findings (unvalidated input in auth flow)
- High: 5 findings (missing rate limiting)

### Performance Summary

- N+1 queries: 3 routes (/api/users, /api/orders, /api/dashboard)
- Missing indexes: 2 tables (users.email, orders.status)

### Launch Blockers

- Unverified high-risk permission rule in `permissions.md`

Integration Within the PM Skills Ecosystem

The AI Shipping Kit occupies slot 9 in the PM Skills suite, as documented in the repository's root README.md and CLAUDE.md files. It serves as the final "hand-off" stage in the product development lifecycle.

While other plugins in the ecosystem handle product strategy, user discovery, go-to-market planning, execution management, and data analytics, the AI Shipping Kit specifically addresses the deployment boundary. It ensures that code generated through AI-assisted development cycles is not only functional but also documented, auditable, and test-validated before reaching production environments.

This positioning makes the kit essential for teams practicing rapid AI generation—where the speed of "vibe coding" requires structured validation to maintain enterprise engineering standards.

Key Source Files and Implementation Details

The following files define the AI Shipping Kit's behavior and integration points:

File	Purpose
pm-ai-shipping/README.md	High-level overview and usage instructions for the plugin
pm-ai-shipping/.claude-plugin/plugin.json	Plugin metadata, version, and command registration
pm-ai-shipping/skills/shipping-artifacts/SKILL.md	Documentation requirements and artifact specifications
pm-ai-shipping/skills/intended-vs-implemented/SKILL.md	Intent-validation methodology and gap-analysis rules
pm-ai-shipping/commands/ship-check.md	End-to-end orchestration logic and packet formatting
pm-ai-shipping/commands/document-app.md	Code-to-documentation generation algorithms
pm-ai-shipping/commands/derive-tests.md	Test coverage analysis and gap identification
README.md (repo root)	Ecosystem positioning (slot 9) and plugin index
CLAUDE.md (repo root)	Cross-reference table mapping all plugins to their purposes

Summary

• The AI Shipping Kit (pm-ai-shipping) is a plugin module in phuryn/pm-skills that creates reviewer-ready shipping packets for AI-generated applications.
• It provides five core commands (/ship-check, /document-app, /derive-tests, /security-audit-static, /performance-audit-static) that can run independently or as a unified workflow.
• The kit relies on two specialist skills (shipping-artifacts and intended-vs-implemented) to enforce documentation standards and validate implementation accuracy against intent.
• The output is a shipping packet containing documentation inventory, security/performance audits, and launch blockers designed for human sign-off and future AI context.
• As slot 9 in the PM Skills ecosystem, it serves as the final quality gate between AI-assisted development and production deployment.

Frequently Asked Questions

What exactly is a "shipping packet" in the AI Shipping Kit?

A shipping packet is a comprehensive Markdown document generated by the /ship-check command that aggregates all validation results into a single reviewable artifact. According to the source code in commands/ship-check.md, it includes a documentation inventory checklist, security and performance audit summaries, test-coverage mappings, identified launch blockers, and actionable next steps. This packet is designed for human reviewer sign-off and provides structured context for future AI development cycles.

How does the AI Shipping Kit differ from traditional code review tools?

Unlike generic linting or static analysis tools, the AI Shipping Kit is documentation-centric and intent-aware. It cross-references actual code against explicitly documented architecture, permissions, and business rules via the intended-vs-implemented skill rather than just checking for syntax errors or common vulnerabilities. Additionally, it produces a standardized shipping packet format specifically designed for product manager and founder review, bridging the gap between engineering implementation and business requirements.

Can I run individual commands without executing the full ship-check sequence?

Yes. While /ship-check orchestrates the complete workflow defined in commands/ship-check.md, each command operates independently. You can run /pm-ai-shipping:document-app to generate documentation without performing security audits, or run /pm-ai-shipping:security-audit-static in isolation to check for vulnerabilities. This modular approach allows teams to integrate specific validation steps into existing CI/CD pipelines or development workflows without running the entire sequence.

Where does the AI Shipping Kit fit in the product management workflow?

The kit occupies the final pre-deployment stage (slot 9) of the PM Skills ecosystem, as listed in the root README.md and CLAUDE.md. It follows strategy, discovery, design, and implementation phases handled by other plugins. Specifically, it bridges the gap between AI-assisted coding (often called "vibe coding") and production release, ensuring that rapidly generated code meets documentation, security, and performance standards before shipping to users.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}