By RostK
Read-only researcher agent that finds grounded answers inside your codebase or on the web and returns cited, structured reports — never writes or mutates anything.
A Claude Code plugin marketplace — a catalog that distributes plugins (skills, commands, agents, hooks, MCP servers) to Claude Code users.
my-monkeys/
├── .claude-plugin/
│ └── marketplace.json # marketplace catalog (required, repo root)
└── plugins/
└── example-plugin/ # template folder — one directory per plugin
├── .claude-plugin/
│ └── plugin.json # plugin manifest (name, description, version)
├── skills/ # <skill-name>/SKILL.md
├── commands/ # flat <command-name>.md slash commands
└── agents/ # <agent-name>.md subagents
The
example-pluginfolders are placeholders (currently.gitkeep).marketplace.jsonships with an emptyplugins: []until you register your first plugin.
Create the manifest plugins/<name>/.claude-plugin/plugin.json:
{
"name": "<name>",
"description": "What this plugin does",
"version": "1.0.0"
}
Add at least one component, e.g. a skill at plugins/<name>/skills/<skill>/SKILL.md:
---
description: Short description of what the skill does
---
Instructions Claude follows when this skill runs.
Register it in .claude-plugin/marketplace.json (relative to pluginRoot: "./plugins"):
"plugins": [
{
"name": "<name>",
"source": "<name>",
"description": "What this plugin does",
"version": "1.0.0"
}
]
claude plugin validate . # validate marketplace.json + plugin manifests
claude plugin marketplace add ./ # register this marketplace locally
claude plugin install <name>@my-monkeys # install a plugin from it
Helper scripts live in scripts/ (Bash — run in Git Bash on Windows, or any shell on macOS/Linux/CI). They use a semver git-tag model: each release is an annotated tag vX.Y.Z, and a rollback restores a previous tag as a new forward commit (no force-push).
# Cut a release: validate the marketplace, tag vX.Y.Z, push.
scripts/release.sh 1.0.0
# Also bump a specific plugin's version before tagging (needs jq):
scripts/release.sh 1.1.0 --plugin example-plugin
# Preview without changing anything:
scripts/release.sh 2.0.0 --dry-run
# Roll back to a known-good tag (safe, non-destructive):
scripts/rollback.sh v1.0.0
Both scripts run claude plugin validate . first, require a clean tree on main, and accept --no-push, --dry-run, and -y. Users receive changes when they run /plugin marketplace update. See scripts/release.sh --help / scripts/rollback.sh --help for all options.
Prerequisite: configure git identity once (
git config user.name/user.email) so tags and commits can be created.
name (marketplace and plugins) must be kebab-case (lowercase + hyphens).name is an immutable slug once published — change the UI label via displayName, never by renaming name.source paths only resolve when the marketplace is added via git or a local path, not via a direct URL to marketplace.json.version (here or in plugin.json) to treat every git commit as a new version; set it and bump on each release to pin.Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub rostk/my-monkeys --plugin research-toolsSpec-Driven Development engineering workflow: spec-creator to implementation-planner to run-plan (implementer + plan-verifier + architecture-reviewer) to retro, with durable per-step harness telemetry.
Generalized architecture-reviewer agent — read-only structural / topology review of a diff, module, or branch, tiering findings Violation/Smell/Nit with path:line citations.
Reusable engineering skills — React, Next.js, Fastify, onion & frontend architecture, testing, security, TypeScript, Zod, Drizzle ORM, PostgreSQL, and engineering-insights capture. The paved-path base other plugins depend on.
Complete creative writing suite with 10 specialized agents covering the full writing process: research gathering, character development, story architecture, world-building, dialogue coaching, editing/review, outlining, content strategy, believability auditing, and prose style/voice analysis. Includes genre-specific guides, templates, and quality checklists.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification
Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, plus codex, antigravity, and grok CLIs when installed) to get diverse perspectives on coding problems
Use this agent when creating user interfaces, designing components, building design systems, or improving visual aesthetics. This agent specializes in creating beautiful, functional interfaces that can be implemented quickly within 6-day sprints. Examples:\n\n<example>\nContext: Starting a new app or feature design
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review