A curated Claude Code plugin marketplace — opinionated, production-ready skills for AI-assisted work.
Quick Start · Plugins · Installation · Roadmap · Maintainer docs
A personal marketplace of Claude Code plugins by Fernando Bertholdo, focused on craft, rigor, and practical excellence. Every plugin is self-contained, documented, and installable in seconds through Claude Code's native plugin system.
Why "nbstack": NB is what my friends call me (Nando Bertholdo), and the name follows Garry Tan's gstack and Lauren Tan's pstack: one person's working stack, published. Until October 2026 this marketplace was called 4 Successful AI Life; GitHub redirects the old links. If you added it under the old name, run /plugin marketplace remove 4-successful-ai-life, add it again with the command below and reinstall the plugins you use.
Inside any Claude Code session, add the marketplace and install the plugins you want:
/plugin marketplace add fernando-bertholdo/nbstack
/plugin install ui-excellence@nbstack
/plugin install smart-session-rename-cc@nbstack
/plugin install prompt-master@nbstack
/plugin install generate-session-prompt@nbstack
/plugin install enhanced-planning@nbstack
/plugin install multica-ops@nbstack
/plugin install cmux-hibernate@nbstack
/plugin install decisoes@nbstack
/reload-plugins
Drop any line you don't want — every plugin is independent. After reload, skills are invocable as slash commands (/ui-excellence:accessibility, /smart-rename, /generate-session-prompt, /enhanced-planning); prompt-master is model-invoked and needs no command — just ask Claude to work on a prompt. Other install paths (local clone, single-plugin test, permanent opt-in) are in Installation.
| Plugin | Version | What it does |
|---|---|---|
| ui-excellence | 1.0.0-alpha.4 |
UI/UX skill bundle with a triage coordinator — visual design, typography, accessibility, usability audits, CRO, microinteractions, and engagement loops. |
| smart-session-rename-cc | 1.5.0 |
Auto-names your Claude Code sessions from the work you actually do, via a Stop hook + a Haiku-generated title. |
| prompt-master | 1.0.1+upstream-1.8.0 |
Generates optimized prompts for any AI tool (LLMs, image/video AI, coding agents). Activates only on explicit prompt-engineering requests. |
| generate-session-prompt | 1.1.1+upstream-4.0.0 |
Generates a handoff prompt to resume work in a new session — for long sessions, pauses, or tool switches. |
| enhanced-planning | 1.1.1+upstream-2.0.0 |
Adds structural guardrails to implementation plans — human checkpoints, risk registry, decision locks, multi-session protocol, and Codex review. Complements writing-plans. |
| multica-ops | 0.2.12 |
Companion to Multica's official CLI skill — desktop-app binary discovery, the bridged-session route, the shapes and limits the reference omits, and the constraints of a governed board. |
| cmux-hibernate | 0.2.0 |
Hibernates the Claude Code sessions open across cmux workspaces before you quit cmux, so they come back one at a time, as you open each tab, instead of all at once. |
| decisoes | 0.1.1 |
A Claude Code mod that collects the decisions only you can make in each session, lets you answer them with buttons, and takes you to the other cmux sessions waiting on you. Texts in Portuguese. |
🎨 ui-excellence — UI/UX craft, framework-agnostic
13 skills (12 specialists + 1 triage coordinator) for building and auditing web interfaces, grouped into Foundations (animation-motion, visual-polish, web-standards, accessibility), Systems (refactoring, typography), Audit (heuristics, cro), Interaction (microinteractions), and Behavior (hooked, retention, copy).
Foundations are original; the Systems, Audit, Interaction, and Behavior groups are adapted from wondelai/skills (MIT, with attribution — see the plugin LICENSE).
Invoke any skill directly when you need targeted guidance:
/ui-excellence:visual-polish # spacing, shadows, optical alignment
/ui-excellence:accessibility # WCAG 2.1 AA — keyboard, ARIA, contrast
/ui-excellence:heuristics # Nielsen + Krug usability audit
/ui-excellence:cro # conversion-rate optimization audit
…or let the _coordinator triage your task and route to the right specialist. → Plugin README
✍️ smart-session-rename-cc — never hunt through quirky-blue-elephant again
A Stop hook scores work-density deterministically and asks Haiku for a structured domain: clause title once meaningful work has accumulated — so a random quirky-blue-elephant becomes auth: add rate limiting, tests. It stays quiet during Q&A and only spends a call when real code is happening.
Seven /smart-rename subcommands keep you in control:
/smart-rename # suggest a title now
/smart-rename explain # show the current state snapshot
/smart-rename <name> # anchor a domain (e.g. /smart-rename billing)
/smart-rename freeze # pause auto-rename (unfreeze to resume)
/smart-rename force # override the throttle / reset the circuit breaker
/smart-rename unanchor # clear the domain anchor
💬 prompt-master — optimized prompts for any AI tool
Model-invoked — no slash command. It activates automatically when you explicitly ask Claude to write, fix, improve, or adapt a prompt for a specific AI tool, and stays out of the way during general conversation or coding. Just ask:
"Improve this Midjourney prompt: a cabin in the woods at dusk…"
"Write a system prompt for a coding agent that reviews pull requests."
Vendored from nidhinjs/prompt-master and synced weekly. → Plugin README
🔄 generate-session-prompt — resume work without losing context
Generates a structured handoff prompt so you can pick up in a fresh session after a long run, a pause, or a tool switch. Dual-mode: opinionated when a .planning/ directory exists in the project root, generic otherwise — so it's useful in any repo.
/generate-session-prompt # mode auto-detected
/generate-session-prompt detailed # detail level: brief | standard | detailed
Vendored from tech-product-template and synced weekly. → Plugin README
🧭 enhanced-planning — structural guardrails for implementation plans
Adds structural guardrails to a plan before you write it: human checkpoints, a risk registry, named guardrails (G-*), decision locks, a multi-session continuity protocol, and a Codex review pass. It complements writing-plans (from superpowers) — enhanced-planning shapes the plan's structure, then writing-plans fills in the content.
Use it when a task spans 3+ PRs or multiple sessions, touches stakeholder-visible output, or risks drift between components.
/enhanced-planning checkout-v2 # guardrails for a unit of work
/enhanced-planning # guardrails, no specific label
It originates from a milestone/detour planning framework but works standalone — a "Standalone usage" preamble in the skill maps the jargon and marks companion skills as optional. Pairs with writing-plans and /codex:rescue (OpenAI's official Codex plugin for Claude Code).
Vendored from tech-product-template and synced weekly. → Plugin README
🗂️ multica-ops — the layer around Multica's official CLI skill
Multica publishes the canonical CLI skill at multica-ai/multica-cli, and it is well maintained — a linter checks its documented commands against the real binary daily. Install it: it is the command reference. This plugin is deliberately not that, and restates none of it.
What it adds: finding the binary in a desktop-app install (where command -v multica returns nothing and the reference's first step fails); reaching the CLI from a Cowork/claude.ai session bridged to that machine; the zsh word-splitting bugs behind the mt() wrapper; the JSON shapes and cost flags the reference does not state (--limit caps at 100, --fields trims context, issue children is not an array, issue timeline exists); the platform limits (workspace isolation, artifacts disabled inside agent tasks, agents consume MCP via agent update --mcp-config); and — the part that only matters if you work this way — the constraints of a board governed by a repository that versions its agent instructions, skills, context and project descriptions: editing is not deploying, ids come from the manifest, drift is measured before state is claimed, and deployment is a human act.
/multica-ops:board
Point MULTICA_PLAYBOOK at that repository and the skill defers to it. Nothing board-specific ships here: no workspace ids, no project names, no machine paths. → Plugin README
💤 cmux-hibernate — quit cmux without losing, or reloading, every session
Dozens of tabs running Claude Code cost memory, and quitting cmux is risky: the resume binding it keeps records an intention, not the live process, and reopening starts every session at once. hibernate.py reads the state from the processes (CMUX_SURFACE_ID, lsof), rewrites each binding with the right directory and auto_resume: false, and writes a dated snapshot; after you reopen, each tab wakes when you open it, and wake.py checks or rebuilds what is missing. --surface <tab> hibernates a single tab, including the one the command runs from.
/cmux-hibernate:cmux-hibernate
⚑ decisoes — the decisions waiting on you, one click away
A mod — a plugin of function hooks, with a band above the prompt and a pane of its own. The agent gets a registrar_decisao tool and an instruction to use it instead of asking in prose; lines its answer flags with 🔴 (or "aguardo seu aval") become decisions too. The band shows how many decisions wait and for how long; the pane takes one at a time, with numbered options, a free answer, and "adiar", and sends the answers back as one turn once the last pending one is answered. The band also lists the other sessions waiting on you and takes you to their cmux tab, or resumes the session in a new tab when the old one closed; two buttons ask the session to land before you close the machine. Texts in Portuguese.
/decisoes
The Quick Start above is the recommended path — install from GitHub. Other ways to install:
From a local clone (for development)
Clone the repository and add the local path as a marketplace:
git clone https://github.com/fernando-bertholdo/nbstack.git
cd nbstackThen, inside a Claude Code session launched from the clone directory:
/plugin marketplace add ./
/plugin install ui-excellence@nbstack
/reload-plugins
Useful when you want to hack on a plugin locally before pushing changes.
Single-plugin test via --plugin-dir
Bypass the marketplace entirely and load one plugin directly:
claude --plugin-dir ./plugins/ui-excellenceHandy for validating a plugin in isolation without touching your global marketplace registry.
Permanent opt-in via settings.json
To auto-install in a project, add to .claude/settings.json:
{
"extraKnownMarketplaces": {
"nbstack": {
"source": {
"source": "github",
"repo": "fernando-bertholdo/nbstack"
}
}
},
"enabledPlugins": {
"ui-excellence@nbstack": true,
"smart-session-rename-cc@nbstack": true,
"prompt-master@nbstack": true,
"generate-session-prompt@nbstack": true,
"enhanced-planning@nbstack": true,
"multica-ops@nbstack": true,
"cmux-hibernate@nbstack": true,
"decisoes@nbstack": true
}
}Claude Code prompts you to trust the marketplace on first open, then keeps the plugins available across sessions.
ui-excellencev1.0.0 — graduate fromalpha: finalize the coordinator's path-aware auto-loading and integration cleanup.smart-session-rename-ccv1.5.1 — fix two CLI-path bugs surfaced during Level 4 testing (state empty-write under degenerate transcripts;/smart-rename forcestate divergence).planning-suite— bundle the milestone/detour planning family (init-milestone,validate-dor/dod,archive-initiative, …) around the already-shippedenhanced-planning, so its companion-skill references resolve within one plugin.- Future plugins — focused bundles (sync, design-sprint workflows) extracted from long-running real-world use.
MIT — see LICENSE. Individual plugins may carry additional attribution in their own LICENSE files when they incorporate third-party content.
This marketplace is maintained for personal and project-specific use. External contributions aren't actively solicited, but issues and discussion are welcome on the issue tracker.
- How plugins are built, vendored, versioned, and synced →
CONTRIBUTING.md - Repo conventions & the anti-drift checklist →
CLAUDE.md
If a plugin here saves you time, a ⭐ helps others discover the marketplace.