marketplace

A personal marketplace of Claude Code plugins and skills, structured so that any Claude Code session can install one of these extensions and use it immediately.

This repository is intended to be consumed by Claude Code (and the Claude Agent SDK where applicable). It is not a runnable application on its own.

Layout

marketplace/
├── .claude-plugin/
│   └── marketplace.json         marketplace manifest: lists installable plugins
├── CLAUDE.md                    project guidance loaded into every session
├── LICENSE
└── plugins/
    ├── web-scaffold/            scaffolds a Next.js + FastAPI website
    ├── business-requirements/   senior BA for AI-augmented quality intelligence work
    └── logical-consistency/     internal logic / fallacy / contradiction audit for documents

A plugin is a self-contained directory under plugins/ with its own .claude-plugin/plugin.json and a skills/<skill-name>/SKILL.md for each skill it exposes.

Installing a plugin

In a Claude Code session, install plugins from this marketplace using the standard plugin install flow (point your session at this repo). Each plugin then becomes available in that session.

Plugins

web-scaffold

Status: 0.1.0

A plugin that scaffolds a new website project from scratch. It ships one skill:

Skill: create-website

Builds a working repository for a marketing/content website with a Next.js frontend, a FastAPI backend, and an AWS deployment story.

When to invoke. Ask Claude to "scaffold a new website", "create a marketing site", or "bootstrap a Next.js + FastAPI project". The skill triggers from natural-language prompts of that shape. It does not trigger for edits to an existing site, single-page generation, SaaS dashboards, or non-web projects.

What it generates.

• Frontend — Next.js 16 (App Router), React 19, TypeScript, Tailwind v4, MDX. Starter pages: /, /about, /blog, /news, /contact, plus not-found, error, sitemap, robots, an Open Graph image, and a web manifest. Vitest tests for lib/ and the contact form. ESLint, Prettier, .editorconfig.
• Backend — FastAPI managed with pdm. Endpoints: GET /health and POST /contact. The contact endpoint is rate-limited and includes a honeypot. Pytest covers the happy path, missing fields, and the honeypot. Multi-stage Dockerfile suited to AWS App Runner.
• Local workflow — Makefile with parallel dev (frontend on :3000, backend on :8000), plus setup, configure, lint, typecheck, test, ci, docker-up, docker-down, build, deploy-check. docker-compose.yml with an optional commented-out Postgres service.
• Deployment — amplify.yml configured for the monorepo layout (appRoot: frontend). References describe AWS Amplify Hosting for the frontend, AWS App Runner for the backend, and Route 53 + ACM for the domain and HTTPS.
• Optional analytics — A small Analytics component routes to Plausible, PostHog, or nothing, driven by NEXT_PUBLIC_ANALYTICS_PROVIDER.
• Project docs — README.md, FEATURES.md, TODO.md, CHANGELOG.md, plus docs/SITE.md where the user describes what the site is about (read by the skill on subsequent runs to inform copy).

Prompted fields. The skill asks for site_name, company, domain, tagline, description, contact_email, github_repo, accent_color, and analytics. The backend domain defaults to api.<domain>. Full list and validation rules: content-prompts.md.

Deeper documentation.

• Plugin README
• SKILL.md — full skill instructions, placeholder rules, file manifest, re-run behavior, example transcript
• amplify-deploy.md — Amplify + Route 53 + HTTPS for the frontend
• fastapi-backend.md — App Runner deploy for the backend, plus ECS Fargate and Lambda alternatives
• TODO.md — quality-reviewed follow-ups

business-requirements

Status: 0.1.0

A plugin that turns rough product ideas into structured, testable, implementation-ready business requirements. It ships one skill, one slash command, and one sub-agent.

Skill: analyze-requirements

A senior Business Requirements Analyst voice with deep expertise in software quality engineering, AI-assisted and agentic automation, document intelligence platforms, and SaaS product design.

When to invoke. Ask Claude to "analyze requirements for X", "write a BRD for…", "what should this feature do", "turn this rough idea into requirements", "draft user stories and acceptance criteria for…", or "refine the scope of…". The skill does not trigger for pure implementation help on settled features, code review, or non-product analysis.

What it produces. A Business Requirements Analysis with up to 18 sections: executive summary, business goals, users and stakeholders, problem statement, scope (in/out/future), key use cases, functional requirements (IDs + MoSCoW priority), non-functional requirements, data requirements, AI and agentic automation requirements, integrations, reporting and analytics, Given/When/Then acceptance criteria, risks and mitigations, assumptions, open questions, MVP recommendation, and roadmap suggestions.

Two layered modes (on by default).

• Product Strategy Mode — opinionated calls on differentiating vs. commodity features, MVP scope cuts, real AI value vs. demoware, deterministic vs. agentic boundaries.
• Implementation Handoff Mode — engineering-ready appendix with epic, user stories, suggested API endpoints, data model changes, UI components, backend services, test cases, observability requirements, security considerations, and a Definition of Done checklist.

Disable either mid-conversation with "skip strategy" or "no handoff format."

Default lens. Unless the user specifies otherwise, the skill assumes the work serves an AI-Augmented Quality Intelligence Platform and uses the preferred vocabulary (Quality Intelligence Platform, Software Knowledge Model, Quality Signal, Source-grounded recommendation, Human-in-the-loop review, etc.).

Slash command: /brd <topic>

Direct invocation of the analysis. Produces a first-pass BRD using the canonical structure, with both default modes applied.

Sub-agent: business-analyst

Delegate BA work from a main agent when you need a senior BA voice on a focused question without consuming the main context window.

Deeper documentation.

• Plugin README
• SKILL.md — full skill instructions and trigger rules
• output-template.md — canonical section-by-section output
• requirement-quality-rules.md — how to write good requirements
• agentic-checklist.md — questions every AI/agent feature must answer
• product-strategy-mode.md
• implementation-handoff-mode.md
• vocabulary.md — preferred Quality Intelligence Platform terms
• platform-context.md — default platform inputs, outputs, agent workflows

logical-consistency

Status: 0.1.0

A plugin that audits a document — or a set of documents — for internal logical consistency: contradictions, common logical fallacies, undefined or equivocated terms, and invalid inference forms. Works on software requirements (BRD, SRS, PRD, spec), design documents and RFCs, essays and blog posts, policies, and contracts. It ships one skill, one slash command, and one sub-agent.

Skill: check-logical-consistency

A senior logic and argument auditor focused on whether a document reasons soundly from its own stated premises. It is not a fact-checker against the outside world, not a copy editor, and not a code reviewer.

When to invoke. Ask Claude to "check this for consistency", "audit the logic of…", "find contradictions in…", "review this BRD/spec/essay for logical issues", "is this argument sound", or "are these requirements consistent". The skill does not trigger for grammar or style edits, external fact-checking, opinion review, or pure code review.

What it produces. A Logical Consistency Report with ten sections: summary verdict, documents reviewed, central claims extracted, contradictions, logical fallacies, informal logic issues (undefined terms, terminological drift, equivocation setup, missing premises, unsupported claims, scope drift, numerical inconsistencies, internal citation drift), inference errors, severity summary (Blocker / Major / Minor / Nit), recommended repair order, and out-of-scope items noted.

Each finding contains ID, type, location, verbatim quote, a one- or two-sentence explanation, a concrete repair, and a severity rating.

Three analysis passes.

• Formal logic — direct contradictions, incompatible constraints, quantifier errors, modal and deontic conflicts, temporal cycles, definitional contradictions, invalid inference forms.
• Logical fallacies — equivocation, begging the question, false dichotomy, hasty generalization, appeals to authority / popularity / nature / tradition, slippery slope, straw man, composition / division, post hoc, no true Scotsman, cherry picking, appeal to ignorance, loaded question, genetic fallacy, and more.
• Informal logic — undefined or under-defined load-bearing terms, terminological drift, equivocation setup, missing premises, unsupported claims, scope drift, numerical inconsistencies, internal citation drift.

Operating principle. The skill applies the principle of charity before flagging any finding and refuses to invent contradictions to look thorough. A clean document earns a short "no significant findings" report.

Slash command: /logic-check <input>

Direct invocation. The input can be one or more file paths or pasted prose. Produces the report inline unless the user asks to save it.

Sub-agent: logic-auditor

Delegate a logical-consistency review from a main agent when you need an audit on a substantial document without consuming the main context window.

Deeper documentation.

• Plugin README
• SKILL.md — full skill instructions and trigger rules
• formal-logic.md — propositional, quantifier, modal, deontic, temporal, and inference checks
• fallacies.md — catalog of common fallacies with detection patterns and examples
• informal-logic.md — undefined terms, drift, missing premises, scope and citation issues
• document-types.md — emphasis per document type
• output-template.md — canonical report structure

Conventions

Repository-wide conventions are defined in CLAUDE.md and are followed by every plugin and skill here:

• pdm for Python package management.
• Make targets for setup, run, lint, test, build, and Docker workflows.
• docker-compose for any local databases or system-level tooling.
• Latest stable APIs and libraries.
• No emojis in code, comments, logs, UI, or generated docs.
• Standard technical writing in documentation.
• Each project maintains CHANGELOG.md, FEATURES.md, and TODO.md.

Status

Early. Structure, conventions, and additional plugins will be added incrementally. Nothing in this repository should be considered stable yet.

License

See LICENSE.