Convert shipwright into curated plugin marketplace

.claude-plugin/marketplace.json

@@ -0,0 +1,21 @@
+{
+  "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
+  "name": "shipwright-marketplace",
+  "description": "RAI's curated Claude Code plugin marketplace",
+  "owner": {
+    "name": "Owais Mohamed",
+    "email": "owais.mohamed@relational.ai"
+  },
+  "plugins": [
+    {
+      "name": "dockyard",
+      "description": "Standalone skills and commands for engineering workflows — brownfield analysis, observability, debugging, codebase profiling",
+      "source": "./plugins/dockyard"
+    },
+    {
+      "name": "shipwright",
+      "description": "Orchestrated agentic development framework — TDD-enforced bug fix workflows with triage, implementation, review, and validation agents",
+      "source": "./plugins/shipwright"
+    }
+  ]
+}

CLAUDE.md

@@ -0,0 +1,45 @@
+# CLAUDE.md
+
+This is **shipwright-marketplace** — RAI's curated Claude Code plugin marketplace.
+
+## Structure
+
+```
+plugins/
+├── dockyard/       ← standalone skills, commands, agents
+└── shipwright/     ← orchestrated workflows (depends on dockyard)
+```
+
+- Marketplace registry: `.claude-plugin/marketplace.json`
+- Plugin manifests: `plugins/<name>/.claude-plugin/plugin.json`
+- Smoke tests: `plugins/dockyard/tests/smoke/` (validates both plugins)
+
+## Plugins
+
+**Dockyard** — standalone tools that work without orchestration:
+- Skills: `brownfield-analysis`, `code-review`, `review-and-submit`, `observability`
+- Commands: `codebase-analyze`, `code-review`, `doc-digest`, `investigate`, `review-and-submit`, `feedback`
+- Agents: `doc-digest`
+
+**Shipwright** — orchestrated bug-fix pipeline (requires dockyard):
+- Commands: `shipwright`, `feedback`
+- Internal agents: `triage`, `implementer`, `reviewer`, `validator`
+- Internal skills: `tdd`, `systematic-debugging`, `verification-before-completion`, `anti-rationalization`, `decision-categorization`
+- Dependency enforced via SessionStart hook (`plugins/shipwright/hooks/check-dockyard.sh`)
+
+## Testing
+
+```bash
+bash plugins/dockyard/tests/smoke/run-all.sh
+```
+
+Runs 4 suites: structure, skills, agents, commands. Validates both plugins.
+
+## Key Conventions
+
+- Commands use YAML frontmatter with `description:` and optional `argument-hint:`
+- Skills live in `skills/<name>/SKILL.md`
+- Agents are markdown files in `agents/` (dockyard) or `internal/agents/` (shipwright)
+- Cross-plugin skill references use `dockyard:<skill-name>` notation
+- `${CLAUDE_PLUGIN_ROOT}` works in hooks.json but NOT in markdown files
+- Bump `version` in plugin.json for users to receive updates (cache is keyed on version)

CODEOWNERS

@@ -0,0 +1,6 @@
+# Marketplace registry -- plugin additions require CODEOWNERS approval
+/.claude-plugin/marketplace.json @omohamed-rai
+
+# Plugin-level ownership
+/plugins/dockyard/ @omohamed-rai
+/plugins/shipwright/ @omohamed-rai

CONTRIBUTING.md

@@ -0,0 +1,43 @@
+# Contributing to Shipwright Marketplace
+
+Shipwright Marketplace is RAI's curated Claude Code plugin marketplace. It hosts two plugins:
+
+- **dockyard** -- standalone skills and commands (codebase analysis, debugging, doc digest, etc.)
+- **shipwright** -- orchestrated workflows that compose dockyard skills into multi-agent pipelines (triage, implement, review, validate)
+
+## Which Plugin Should I Contribute To?
+
+| If your contribution is... | Add it to |
+|---|---|
+| A standalone skill or command (usable on its own) | `plugins/dockyard/` |
+| An agent, workflow, or orchestration that composes skills | `plugins/shipwright/` |
+
+## How to Contribute a Skill or Agent
+
+1. **Fork and branch** from `main`.
+2. **Add your skill** under the appropriate plugin directory following the required structure:
+   - Skills: `plugins/<plugin>/skills/<skill-name>/SKILL.md`
+   - Agents: `plugins/<plugin>/agents/<agent-name>.md`
+   - Commands: `plugins/<plugin>/commands/<command-name>.md`
+3. **Follow the templates** in `templates/SKILL_TEMPLATE.md` or `templates/AGENT_TEMPLATE.md`.
+4. **Bump the version** in the plugin's `plugin.json`. Claude Code caches plugins by version string, so users will not receive your update unless the version is incremented.
+5. **Open a PR** against `main`.
+
+## Quality Gates
+
+Every PR must pass:
+
+- **Template compliance** -- skill/agent files follow the required format
+- **Smoke test** -- the skill or agent can be invoked without errors
+- **CODEOWNERS review** -- at least one approval from a designated code owner
+
+## Requesting a New Plugin
+
+New plugins require CODEOWNERS approval. To request one:
+
+1. Open an issue describing the plugin's purpose and scope.
+2. A code owner will review and, if approved, add the plugin entry to `marketplace.json`.
+
+## Version Bumps
+
+Claude Code caches installed plugins by version string. If you change any skill, agent, command, or configuration in a plugin, you **must** bump the `version` field in that plugin's `plugin.json`. Without this, users will not see the update until they manually reinstall.

README.md

@@ -1,80 +1,46 @@
-# Shipwright
+# Shipwright Marketplace
 
-Adaptive agentic development framework for engineering teams. Claude Code plugin.
+RAI's curated Claude Code plugin marketplace.
 
-## What It Does
+## Plugins
 
-Shipwright orchestrates a disciplined bug fix workflow: triage, TDD implementation, code review, and validation -- with codebase context and crash recovery.
+| Plugin | Description | Commands |
+|--------|-------------|----------|
+| **dockyard** | Standalone skills & commands | `codebase-analyze`, `code-review`, `doc-digest`, `investigate`, `review-and-submit`, `feedback` |
+| **shipwright** | Orchestrated bug-fix workflows (requires dockyard) | `shipwright`, `feedback` |
 
 ## Installation
 
 ```bash
-# Add the RAI marketplace (one-time setup)
-/plugin marketplace add RelationalAI/claude-plugins
+# 1. Add the marketplace (one-time)
+/plugin marketplace add https://github.com/RelationalAI/shipwright
 
-# Install the beta plugin
-/plugin install shipwright-beta@rai-claude-plugins
-```
-
-> **Stable release** coming soon. For now, Shipwright is available as a beta plugin.
-
-## Usage
+# 2. Install dockyard (standalone skills and commands)
+/plugin install dockyard@shipwright-marketplace
 
-### Orchestrated Workflow
+# 3. (Optional) Install shipwright (orchestrated workflows -- requires dockyard)
+/plugin install shipwright@shipwright-marketplace
 
+# 4. Restart your Claude session to activate
 ```
-/shipwright                                    # Start -- Triage asks what you're working on
-/shipwright fix null pointer on details click   # Start with context
-/shipwright fix bug RAI-9874                    # Start from Jira ticket (requires Atlassian MCP)
-```
-
-**Flow:** Triage -> Implementer -> Reviewer -> Validator
-
-### Standalone Commands
 
-| Command | What it does |
-|---------|-------------|
-| `/shipwright:codebase-analyze` | Full codebase analysis -- writes 7 profile docs to `docs/codebase-profile/` |
-| `/shipwright:doc-digest <path>` | Walk through any document section by section |
-| `/shipwright:debug` | Standalone systematic debugging (4-phase) |
-| `/shipwright:report [description]` | File bugs, feedback, and suggestions on Shipwright |
-| `/promote [comment]` | Cherry-pick skills/agents/commands from beta to stable |
+## Team Setup
 
-## Agents
+Add to your project's `.claude/settings.json`:
 
-| Agent | Role |
-|-------|------|
-| **Triage** | Reads codebase profiles, brainstorms with user, categorizes decisions, confirms tier |
-| **Implementer** | Root cause investigation, TDD, systematic debugging, fix verification |
-| **Reviewer** | Spec compliance review, code quality, approve/challenge/escalate |
-| **Validator** | Full regression testing, fix confirmation |
-| **Doc Digest** | Interactive document walkthrough |
-
-## Skills
-
-| Skill | Purpose | Source |
-|-------|---------|--------|
-| TDD | Test-driven development discipline | [Superpowers](https://github.com/obra/superpowers) |
-| Verification | Evidence before claims | [Superpowers](https://github.com/obra/superpowers) |
-| Systematic Debugging | 4-phase root cause investigation | [Superpowers](https://github.com/obra/superpowers) |
-| Anti-rationalization | Resist shortcuts and "LGTM" | [Superpowers](https://github.com/obra/superpowers) |
-| Decision Categorization | LOCKED/DEFERRED/DISCRETION decisions | [GSD](https://github.com/gsd-build/get-shit-done) |
-| Brownfield Analysis | 7-doc codebase profiling | [GSD](https://github.com/gsd-build/get-shit-done) |
+```json
+{
+  "extraKnownMarketplaces": [
+    "https://github.com/RelationalAI/shipwright"
+  ]
+}
+```
 
-## Design Docs
+## Contributing
 
-- [Design doc](docs/design/shipwright-design-v1.md) -- the full design
-- [Comparison](docs/research/shipwright-vs-others-v1.md) -- how Shipwright compares to Superpowers, GSD, and Beads
-- [Ideas from Beads/GSD](docs/research/shipwright-ideas-from-beads-gsd-v1.md) -- ideas reviewed, adopted, and deferred
-- [M1 Milestone](docs/milestones/m1-tier1-bugfix.md) -- Tier 1 bug fix scope and plan
+See [CONTRIBUTING.md](CONTRIBUTING.md).
 
 ## Attribution
 
-Shipwright builds on the work of:
-
-- **[Superpowers](https://github.com/obra/superpowers)** by Jesse Vincent -- TDD, verification-before-completion, systematic debugging, and anti-rationalization skills. Licensed under Apache 2.0.
-- **[GSD (Get Shit Done)](https://github.com/gsd-build/get-shit-done)** by gsd-build -- Decision categorization and brownfield codebase analysis patterns.
-
-## License
-
-[TBD]
+- **[Superpowers](https://github.com/obra/superpowers)** by Jesse Vincent -- TDD, verification, systematic debugging, anti-rationalization skills. Apache 2.0.
+- **[GSD](https://github.com/gsd-build/get-shit-done)** by gsd-build -- Decision categorization and brownfield analysis patterns.