A curated collection of Claude Code projects, agent workspace blueprints, and related resources — organized by use case. Most patterns here adapt to other agentic AI CLIs and frameworks.
Browse online · Plugins Marketplace · Documentation portal · What are Claude Spaces?
🧭 Start here: Marketplace Guide. This is the suggested way into my Claude Code Plugins Marketplace. Install it (
/plugin install marketplace-guide@danielrosehill) and ask Claude "what's in the marketplace?" or "is there a plugin for X?". It browses every plugin by theme (Israel & Hebrew, homelab, Linux desktop, Claude Code tooling, shopping, writing…), recommends plugins for a task and shows what's new, all read live from the marketplace manifest.
Workspaces by Domain
- Systems Administration
- Productivity & Planning · Legal · Health & Wellbeing · Communications & Writing · Financial Planning · Career · Business · Privacy & Anonymity · Technology & Hardware · Marketing
- Research
- Argument and Perspective Exploration
Configuration & Tooling
Extensions & Scaffolds
- Plugins
- Templates / Scaffolds — recommended way to spin up a new workspace
- Slash Commands
Other
I've been using Claude Code daily for about six months — for development, but also audio editing, legal research, SEO analysis, health documentation, systems administration, and a long tail of non-code use cases. This index is the result: a collection of agent workspaces (repositories structured as self-contained environments for a specific activity) alongside supporting tooling — plugins, context files, MCP servers, and slash commands.
If there's a common thread, it's treating Claude Code less as a coding assistant and more as a general-purpose agent workspace that happens to run in a terminal.
| Type | What it is | Badge |
|---|---|---|
| Agent Workspace | Pre-configured repo using Claude as a conversational UI for a domain-specific workflow |  |
| Template | Forkable starting point you can customize |  |
| Non-Code | Applications beyond software development |  |
More context: the Agent Workspace Model, growth chart, praise
All workspaces in this index follow the same pattern: a Git repository isn't just for code — it can serve as a complete, self-contained workspace for any activity. Each workspace uses a defined folder structure, a CLAUDE.md for agent instructions, slash commands, MCP configurations, and subagent definitions to create a purpose-built environment.
This pattern has been applied to everything from sysadmin and remote server management to legal research, health documentation, and financial planning — domains that have nothing to do with software development.
Primitives globally, scaffolds per-cluster. The tooling has been consolidated into 28 cluster plugins (see Plugins) — each one ships the domain primitives globally (commands, skills, agents for that cluster) and provisions a project scaffold on demand. So rather than forking a separate template repo for each new workflow, you install the relevant cluster plugin once and ask Claude Code to provision a scaffold wherever you need one.
"This is either the work of a prolific genius, or a very clever bot (or both), although it hardly matters because the quality is so good - an index of 75+ Claude Code repositories published by the author... CMS, system design, deep research, IoT, agentic workflows, server management, personal health... If you spot the lie, let me know, otherwise please check these out."
For the record: I'm a real human (danielrosehill.com). The repos and workspaces in this index are generated with Claude Code but human-designed and refined.
Projects involving using Claude for local or remote systems administration as distinct from development-related projects.
See also: The Claude Code Sysadmin Workspaces Index is a dedicated sub-index for all sysadmin workspace templates.
Workspace for managing bash aliases with YADM synchronization support.
Collection of bash aliases for common Claude Code operations on Linux.
GUI concept for Claude-driven LAN management, covering device discovery and network operations.
Planning workspace for managing a Linux desktop with an agentic CLI, running raw notes through a refine/analyse/workflow pipeline.
Multi-plugin installation and setup utility for orchestrating everyday Linux desktop use through Claude Code.
Template using Claude Code as an Ansible stand-in for repetitive environment setup and supervision.
Concept for deploying Claude Code into recovery shell environments for AI-assisted system repair.
Raspberry Pi jump host running the Claude Code CLI over Cloudflare Access and Tailscale as an emergency recovery surface for a Linux workstation.
Custom GRUB boot entry integrating Claude CLI into Linux system recovery workflows.
Projects specifically targeting KDE Plasma desktop integration and Linux desktop workflows with Claude Code.
KDE Dolphin right-click context menu actions (service menus) for launching Claude Code in various Konsole window layouts, including single terminal, split panes, and multi-instance grids.
Launch utility for pairing Claude Code with Konsole on KDE Plasma.
Early standalone Claude Code plugin with slash commands for managing Conda environments and packages.
Per-machine, local-first SQLite log of everything the agent installed, fixed or diagnosed on a Linux desktop, so later sessions can query history before acting. Pairs with linux-desktop-plugin.
KDE Dolphin right-click service menus that open the selected directory in Claude Code or Codex, in normal or permissive mode.
Plan for a KDE-native runner that starts and supervises Claude Code sessions from the desktop.
Claude Code plugin that turns a LAN into working memory: devices, services, operating systems, ports and topology captured in a greppable knowledge store that grows as you and Claude discover things.
Early standalone Claude Code plugin with slash commands for server monitoring, deployment and administration.
Claude Code plugin for administering a Synology NAS over SSH, with guided first-run setup, persistent context and commands for shares, volumes, mounts and storage monitoring.
Workspaces for decision-making, personal planning, file management, and general-purpose productivity workflows.
See also: The Budgeting, Personal Planning, Career, Purchasing, Shopping, and Ideation & Planning cluster plugins in the Plugins section cover these domains.
Index of templates, skills and demos for agent-driven purchase research, built on decoupling the requirements spec from the current market.
Staged, agent-supported process for evaluating any purchase over about $50: capture the spec, find the requirements, survey the market, then score candidates against the spec.
Codifies the iterative spec-refinement loop used with Claude for builds, hardware purchases and home-lab decisions — draft spec, correction, refinement — as reusable skills.
Model workspace for a Claude shopping research tool customised by shop whitelist and task, with input and output folders.
Demo of Claude Code as a consumer purchasing assistant: a written spec plus retailer catalogue screenshots go in, a ranked recommendation report comes out (worked example: a portable Bluetooth speaker).
Work-in-progress CLI and workspace for importing recordings from a digital voice recorder, organising them into dated folders, classifying them with AI and transcribing them.
Workspaces and templates for legal research, case management, and evidence handling workflows.
See also: The Legal & Investigative cluster plugin in the Plugins section covers this domain — evidence logging, document analysis, redaction, and brief generation.
Claude Code template for building version-controlled containers for legal files and case material.
Workspaces and templates for health documentation, medical visit management, therapy tracking, and health-related research.
See also: The Personal Planning cluster plugin in the Plugins section covers this domain — diary, health, therapy, preparedness, and personal development variants.
Research notebook investigating ADHD medication access, structured for sustained Claude-assisted enquiry.
Subindex of Claude Code templates, plugins and workspaces for mental health work a person does on their own material: organising events, tracking therapy, preparing for appointments.
GitHub template for working through one bounded personal issue with Claude Code — a career situation, a relationship, a bereavement — with sessions, threads and timelines kept in one workspace per issue.
Workspaces and templates for content creation, blog management, writing workflows, and communications strategy.
Pattern spec for writing full-length reference and instructional books with Claude Code, using teams of subagents and the repository as shared memory, through to a print-on-demand paperback.
Per-project workspace skeleton for producing one AI-generated video; the state half of a pair whose operational half is the ai-video-producer plugin (scripting, image-to-video and lip-sync pipelines).
Workspace scaffold implementing the Agentic Book Writing pattern: a book.toml manifest, spec and interview folders, binding prose and graphics style guides, and a status script for subagent-driven drafting.
Automated workflow for sending polished update emails about website changes via Resend MCP.
Personalized text rewriting rules that consolidate into a slash command for refining AI-generated documentation.
Instruction-only agent skills that keep Hebrew, Arabic, Farsi and Urdu text right-to-left on every surface an agent writes to (Gmail, Resend, Google Docs/Sheets/Slides) and verify the setting took. Works in Claude and ChatGPT.
Early standalone Claude Code plugin with slash commands for proofreading, style editing and text transformation.
Workspaces and templates for budgeting, purchasing decisions, and personal finance management.
See also: The Budgeting, Purchasing, and Shopping cluster plugins in the Plugins section cover these domains.
Claude Code workspace for FX pair analysis — historical data, quant and vol modelling, geopolitical context, and Typst PDF reports.
Work-in-progress skills basket for buyer-side Israeli residential mortgage research — track mix, Bank of Israel rules, DSR/LTV limits and bank quirks — before verified skills are folded into the Israel agent-skills plugin.
Work-in-progress skills basket for Israeli property due diligence (Tabu extracts, building files, usage permits, comparables), aimed at non-professional buyers and tenants.
Workspaces and templates for job searching, career planning, and professional development.
See also: The Career cluster plugin in the Plugins section covers this domain — role logging, offer comparison, application tracking, and salary benchmarking.
Workspaces and templates for business planning, idea evaluation, and organizational continuity.
See also: The Ideation & Planning cluster plugin in the Plugins section covers business idea evaluation, decision frameworks, and simulation workflows.
Workspaces and templates for document redaction, identity protection, and PII obfuscation.
See also: The Legal & Investigative cluster plugin in the Plugins section includes redaction and document-obfuscation workflows. For broader system hardening see the Security Checkup plugin.
Assistant configuration for redacting personally identifiable details from documents.
Workspaces for hardware planning, PC builds, and technology procurement.
See also: The Purchasing (includes a tech-procurement variant), Sysadmin & Homelab, and HP5200 Printer plugins in the Plugins section cover these domains.
Dedicated macropad for driving Claude Code — approve, stop, new session, transcript navigation — with a host-driven status LED that lights when Claude is waiting on you.
Ergonomic and geometric research into a display for running a grid of at least eight concurrent Claude Code sessions, measured in text cells rather than inches, plus the software layer for navigating the grid.
Non-code planning workspace for a home media setup: ideas, voice notes and reference material refined into requirements and decisions.
Workspaces for SEO, web analytics, PR monitoring, and media tracking.
See also: The PR & Media Work cluster plugin in the Plugins section covers coverage scanning, press summarisation, response drafting, and comms strategy.
Projects using Claude and agentic systems for deep research, report generation, and information synthesis.
See also: The Research Space cluster plugin in the Plugins section covers deep research, technical research, OSINT, geo-reaction, stack, ecosystem, and competitor research workflows.
Research notes on Agent-to-Agent (A2A) protocols and how they relate to Claude.
Framework for iterative deep research using Claude with voice pipeline and structured outputs.
Agent that generates platform-specific search dorks across Google, Reddit, Twitter/X, and more.
AI-assisted municipal platform for a candidate who doesn't exist: a researched, cited and costed Jerusalem mayoral manifesto built with Claude Code. Sibling of MK Claude.
Research workspace on using Claude and Claude Code for operating system management.
Claude Code plugin for researching, filtering, comparing, and evaluating models on OpenRouter, pulling the live catalogue without an API key.
Public record of build-vs-adopt scans for Claude Code tooling: does a plugin, MCP server or skill already exist for a given requirement, or is it blue sky? Each scan keeps the searches — including the empty ones — against a closed verdict set with dated decay.
Ongoing research notebook, built on the Claude research workspace template, on managing a large personal GitHub fleet: organisation, multi-machine sync, bulk lifecycle operations and launcher UX.
Voting-decision workspace for the October 2026 Knesset election: a voice note of personal priorities becomes a weighted rubric, Claude agents research and score every list on the ballot, and the result ships as ranked PDF reports and an interactive Hugging Face Space.
Instruction-only plugin of Israel-specific research skills — Kol Zchut rights, Home Front Command guidelines, travel warnings, medicines, government offices and opening hours — that runs in Claude and ChatGPT.
Claude-assisted research workspace cataloguing the UX of live voice typing as a keyboard replacement on the desktop: interaction patterns, inference approaches and tool comparisons.
AI-assisted political platform experiment for Israel: a detailed, researched and costed manifesto for a party that doesn't exist, written in public with Claude Code.
GitHub template for AI-assisted research that produces a working Obsidian vault: frontmatter and wikilink conventions are set up so the agent's notes populate the graph, backlinks and properties.
Projects using AI for synthesized debate to explore various perspectives, including policy modeling and analysis.
Simulated conference in which agentic personas (HR leads, CEOs, doctors, artists, sceptics) each deliver a three-minute talk on "AI in 2026", rendered with TTS.
Workspace for challenging personal beliefs through AI-generated counterarguments and rebuttals.
Exploratory model running multiple agents as a panel debate.
Claude Code workspace for finding people who hold specific viewpoints within a policy debate, surfacing voices that keyword search misses.
Projects exploring using Claude and related tooling for personalized user engagement, including through RAG, interviewing methods, and context injection.
GitHub template repository pre-seeded with every AI Dev Workspace Primitive as a stub, with instructions for the user or agent to fill each one in.
October 2025 template for AI-assisted development with a layered CLAUDE.md hierarchy that guides the agent through project stages.
Index of root-level UPPERCASE.md context primitives for AI-assisted repos (SPEC.md, RULES.md, TOOLS.md, SECRETS.md, GROUNDING.md and more), each published as its own repo and injectable via CLAUDE.md/AGENTS.md.
Automation workspace for batch-adding CLAUDE.md files across multiple GitHub repositories.
Pre-configured CLAUDE.md templates for managing different repository types.
Redacted point-in-time dump of /context from a heavily-pluginned Claude Code session, analysing where the context budget actually goes and what is eagerly versus lazily loaded.
Index of the user-level layer of a Claude Code setup, treated as one system — entrypoint, reference context, habits and discovery — covering what loads before any task is known.
Prompt template for verifying the correct Claude model variant at conversation start.
Paste-ready CLAUDE.md / AGENTS.md snippet that gives an agent a spending approval limit: it spends freely on paid APIs below a tunable threshold, and above it must estimate the cost, offer a cheaper route and wait for approval.
Depersonalised reference pattern for organising user-level ~/.claude/ context: a lean top-level CLAUDE.md that routes to topical context/*.md files (system environment, git rules, MCP usage, file organisation, etc.) loaded only when relevant.
Pattern for maintaining separate CLAUDE.md files for developers and for end users of the same repository.
Workflow methodology for separating human-authored context from structured AI agent briefings.
Splits a user-level system prompt into one file per standing habit, assembled into a pasteable block and a JSON index. Ships skills to install a subset into any CLAUDE.md and to reconcile drift when the prompt and the repo disagree.
Working CLAUDE.md for a home directory on an Ubuntu Linux desktop.
Automatically generates contextual CLAUDE.md files across a Linux desktop filesystem.
Annotated corpus of writing samples used as ground truth for a personal writing style, with reusable methodology for building a tone-of-voice reference for AI agents.
Tools for managing public and private CLAUDE.md files with security-focused git configuration.
Pattern for splitting a home-level CLAUDE.md into directives plus on-demand context files, optimised for context handling.
Q&A notes on where context bloat accrues given the current shape of the Claude Code harness and its primitives.
CLAUDE.md snippets and slash commands telling Claude to silently infer around transcription errors from voice typing and stray keystrokes from one-handed or distracted typing.
Components and tooling for multi-agent development and orchestration frameworks.
Design model for an orchestration agent that buffers a human from a batch of working sub-agents — one consolidated brief per cycle, a budgeted override channel, and resumable approval gates.
Pattern for handing tasks between code-generation agents working on different machines through shared handover docs, giving supervising humans an audit trail of what each agent did.
MCP server enabling encrypted peer-to-peer communication between Claude Code instances on localhost or LAN.
Framework for assembling context-optimized multi-agent crews with minimal overlap.
Launchpad for creating standardized workspace templates conforming to the Agent Workspace Model v1.1 spec, with slash commands to generate, validate, and publish new workspaces.
Workspace model for one operator running agentic AI across many repositories all day, borrowed from air traffic control: a voice-first shift manager routes work across concurrent Claude Code sessions. Published as a 26-page typeset document.
Specification for a local proxy that assigns priorities to concurrent Claude Code sessions and shapes their aggregate traffic, so rate limiting degrades background work instead of landing at random. Includes a prior-art survey of the multi-agent proxy and queueing landscape.
Curated toolkit of 74+ Claude Code configurations for development workflows and multi-agent coordination.
Collection of system prompts and configurations for development, operational, and creative tasks.
Curated collection of Claude Code projects and resources organized by category.
Architecture for Claude Code plugins that do sustained work: a public plugin, a private workspace instantiated from a template repo, and a vault at ~/.claude-plugins/<name>/ for user data.
Multimodal workspace for defining Claude Code skills from voice notes, videos, images and text, or written specs.
Idea capture for a meta-plugin holding a personal vault of plugins and MCP servers, activated selectively per project to mitigate eager plugin description loading at user level.
Interactive tool for discovering, selecting, and cloning Claude Workspace templates.
Projects related to Claude and MCP tooling and setup.
Early standalone Claude Code plugin with slash commands for AI tool management and MCP server configuration.
Generator for creating MCP server configuration commands for Claude Code.
Curated index of MCP servers organized into 14+ categories for extending Claude Code.
Home-folder CLAUDE.md snippet giving an agent guidance on which MCP server to reach for.
Template /install-mcp slash command that routes MCP server installations across multiple MetaMCP instances using a tiered preference system.
Guide for instructing AI agents on how to provision and manage MCP server connections according to user-specific preferences, with a tiered decision matrix.
Documentation for integrating Claude Code with MCPM external MCP server manager.
Curated collection of 35+ MCP servers with interactive installer across 15+ categories.
All plugins registered in the danielrosehill marketplace. Install any of these with /plugin install <name>@danielrosehill.
Linux desktop management — auto-profiles the local machine on first run and persists it to user data, then runs system checks, package install/remove, config application, hardware troubleshooting, service/log inspection against that profile.
Register machines and run repeatable Claude-Code-driven security audits over SSH, with timestamped reports and per-machine profiles.
Agent dev workflows: voice notes, screenshots, UI review. Any platform: steer work from voice notes, screenshots and screen recordings; review the running UI; resolve UI friction by options-then-build.
Debugging — capture logs, isolate issue, diagnose error, track bugs, with code/system/issue variants. Includes a KDE hotkey utility for capturing live system bugs.
Bug-ticket workflow for development repos — capture bugs into planning/bugs/, dispatch specialist remediation agents (reproducer, diagnoser, patcher, fix-documenter), document fixes, and ship releases.
Dev-tools — scaffold repos, multi-agent QA review, templatize. Session-handover commands and agent moved to claude-hopper in 1.2.0.
Kaizen for mature codebases: atomic improvements, one commit each, tracked pass by pass from a baseline. Init, intake, step, pass skills.
Kaizen for any project, not just code: small steps, a shared journal, reviews, the agent's Kaizen thoughts as PDF.
Foundational workspace — setup, context management, report parsing, inventory, template discovery, with 6 variants.
Builds and maintains a persistent, portable background context layer about the user — intake interview, ingestion of material they already have, gap analysis, scoped retrieval, maintenance and export. Plain markdown entries in a store the user owns, read through declared scopes and sensitivity levels; explicitly does not use model-managed memory. Ships the Portable Context Contract so issue-scoped workspaces can read it without re-asking who the person is.
Meta-plugin for per-project activation of dormant plugins and MCP servers from a personal vault. Mitigates user-level eager skill description loading.
Backend-agnostic persistent user memory for Claude Code. Ships a save/recall/commit contract with personal/work context routing; defaults to the memory upstream on an MCP gateway, or bring your own memory MCP via a workspace memory-config.md.
Ideation and planning — capture, generate, evaluate, rank, simulate, and plan ideas, with ideation/single-idea-eval/multi-idea-ranking/feature-ideas/simulation/idea-capture variants and Typst PDF deliverables.
Legal and investigative — log evidence, analyze documents, redact, generate briefs, with legal-research/evidence/osint/document-analysis variants.
Modular research task: brief in, PDF + podcast out. Takes a (dictated) brief through a GitHub repo as research notebook, parallel-subagent research with a source log, a Typst PDF report and a two-host educational podcast (Gemini TTS), then Google Drive and email. Every output module is optional and runs alone; the repo can stay a one-off or keep growing in follow-up rounds.
Research — source log, summarize, deep-dive, export, with deep-research/technical/osint/georeaction/stack/ecosystem/competitor/purchasing/general-research-workspace/obsidian-vault variants. The obsidian-vault variant scaffolds the research loop as a working Obsidian vault (committed .obsidian/ config, frontmatter schema, wikilinks, templates, canvas). Includes a 30-agent tech research team for hardware/software stack evaluations (folded in from Claude-Tech-Research-Team).
Check what people are actually saying about a topic, product, or provider by searching curated social-discourse sources (Reddit, Hacker News, Stack Exchange, Trustpilot, YouTube, Lobsters).
AI transparency — document human vs AI contributions, add attribution, audit provenance.
Prompt engineering — craft, eval, catalog, version, search prompts, with library/factory variants.
Drive an AI-generated video project end-to-end: creative brief, model selection, character sheets, script, storyboard, generation pipelines (text-to-image-to-video, voice-to-lip-sync, text-to-video, upscale), clip assembly, and final export. Ships fal.ai/Replicate/MiniMax MCP servers and fal-js + WaveSpeed Python SDK runners.
Audio production — normalize, VAD, transcribe, diarize, podcast assembly, with engineering/podcast/transcript variants.
Media library — catalog, tag, search, sort, dedupe assets.
PR and media monitoring — scan coverage, summarize press, draft responses, comms strategy, with monitoring/response/strategy variants.
Transcription in stages: collect (local, Drive, S3, URL), preprocess (EQ, denoise, silence trim), transcribe (OpenRouter/Gemini Flash, AssemblyAI, Gemini, local Whisper/Parakeet), clean, label speakers, reshape (spec, minutes, to-do, blog, 206 bundled text-transformation prompts), Typst PDF, save. Batch mode runs several files concurrently. Preferences in a plain file.
Video editing, transcoding, video processing, and multimedia workflow automation. Two-tier workspace (index + project), per-user data store, and a growing set of ffmpeg/MLT/Kdenlive primitives.
Content writing — draft, proofread, version, publish, style guides, with writing/blog/opinion/document variants.
Knowledge documentation — index, cross-link, build taxonomy, version docs, with wiki/resource-library/process-docs/experiment-report variants.
Technical documentation — READMEs, reference docs, changelogs, environment docs, with api-reference/code-docs/environment-docs/dev-notebook variants.
Amazon.com marketplace research — verified price, stock, rating, seller and specs read from the listing rather than the search grid; Prime-aware delivery dates checked against the ZIP they were actually rendered for; filtered signed-in search with tested extractors, a marketplace profile and a durable brand allowlist.
Personal budgeting — log transactions, categorize, forecast, track goals, monthly reports.
Career planning — log roles, compare offers, track applications, salary benchmark, with planning/job-search/salary variants.
Personal life planning — log entries, review progress, set goals, with diary/health/family/house-search/preparedness/personal-dev/inbox variants.
Product research and procurement skills with a particular focus on technical specification conformance — does this part actually meet the requirement, and can you show your working. Marketplace-agnostic: spec definition from a vague want or a photograph, annotated search vocabulary, versioned spec-as-code, manufacturer datasheet retrieval, cross-region part-number equivalence, conformance evaluation matrices, bills of materials and RFQ packages. Plus the surrounding process: intake, research, compare, evaluate, recommend, market-check, spec-driven market landscape surveys, live price/delivery scanning through your own VPN egress, preference memory and hardware-rig profiling. Hands off to the marketplace-specific plugins (shopping, israel-shopping, amazon, aliexpress-israel-skills) once a question becomes about a particular shop.
Region-specific consumer shopping primitives — find products, compare local vendors, check availability, and generate ranked purchase recommendations. Ships variant scaffolds per supported region and a provisioning skill for fresh shopping workspaces. Marketplace-specific research (Amazon US, AliExpress Israel) now lives in dedicated per-marketplace plugins; the marketplace-plugins skill is the roster.
Organising therapy reflections — pre/post-session notes, goal tracking, and turning voice-memo transcripts into structured problem summaries. Not therapy: organises notes only. Workspace data lives outside the plugin so the same install survives plugin updates.
Smart home — Home Assistant, Snapcast multi-room audio, Plex media server ops, with HA/audio/media-server variants.
Filesystem organisation — scan, dedupe, cleanup, rename, sort for local directories and Google Drive, with local/gdrive variants. Includes organise-filesystem (modular modes) and super-organise (comprehensive single-pass) skills.
Author and maintain unofficial MCP servers for third-party APIs that don't ship one. Generate a server from the vendor's API docs verified against the live API with your own key, detect upstream drift on a schedule, and smoke-test against real operations. Bound by a committed api-surface.yaml recording what the API actually does versus what its docs claim. Publishes to npm.
Document indoor noise pollution and analyse recordings with ffmpeg, sox and aubio: hums, tones, drone, impacts, isolation from background.
ADB (Android Debug Bridge) operations — onboard a phone, map folders, import media, capture screenshots, manage bloatware with a persistent log, and connect over Wi-Fi with automatic reconnect.
A second spam filter for pseudospam (outreach produced at scale, dressed up as personal): parallel Haiku judges taught by your own Gmail example labels, a review-queue label instead of deletion, a reply queue for genuine mail, plus sender blocks, unsubscribe replies and tracking-pixel/DNS blocking.
Start fresh or hand off: split the pane, launch a new named Claude session and give it the next task, or hand work to a running peer. Payload goes to a local or Google Drive agent-to-agent space and is sent with the built-in SendMessage; this session's Konsole pane then closes. Also asynchronous a2a/ mailbox messages between repos.
Look up current and historical air quality, calculate AQI from raw pollutant readings, and run modelling utilities. Defaults to WAQI with fallbacks to OpenAQ and AireLibre.
AliExpress shopping skills for Israel-based buyers — honest deep search that resolves true SKU prices behind anchored and bundle cards, fast-delivery-to-Israel filter, locale preference with drift guard (default EN/USD), Choice-first search, Israeli-reviews filter, landed cost, findings export (JSON/Markdown/PDF) and the $75 de-minimis across listings and cart.
Skeptical analyst toolkit for long reports — READ/SKIM/SKIP verdicts, structured extraction (arguments, findings, stats, case studies, key snippets), and an opinionated executive summary. Built-in Jaded Report Reader persona that refuses credit for filler.
Audio → branded music video (avp): waveform, logo, karaoke. Install audio-visualizer-python, pull logo/colours/font from a repo into a brand kit, render an MP3 (e.g. from Suno) through JSON templates (waveform, spectrum bars, logo pulse, 9:16), burn in karaoke lyrics from LRC/SRT, and encode small with AV1 by default.
Build, maintain, and audit curated GitHub resource lists (Awesome-style indexes) with AI-driven categorisation, alphabetised tables, and dynamic shields.io badges.
Plan, document, and implement a backup and data-protection strategy for the current project — from architecture discovery through script generation and restore drills.
Bluetooth on Linux from the command line, because the desktop pairing dialogs are unreliable. Diagnose the adapter, scan, pair + trust + connect in one non-interactive bluetoothctl session, manage known devices (connect, disconnect, forget, battery) and reset the stack.
Produce books, long-form research reports, screenplays and audiobooks with teams of subagents. One pipeline (definition, planning, parallel unit drafting, review), then a production stage per type: print handoff, PDF, Fountain, or spoken script + per-chapter TTS + chaptered m4b. Consistency is carried by a style guide, per-unit briefs and a continuity ledger, and a seam reviewer checks every join between units. Successor to book-writing.
Manage BTRFS snapshots via snapper and btrfs from Claude Code — create, list, diff, rollback, and prune subvolume snapshots. Includes /snap and /snap-before commands for one-shot and pre/post-paired snapshots around risky changes.
Evaluate, refine, and develop business ideas through structured lenses — VC review, TAM, B2B/B2C fit, assumptions, objections, hardware feasibility, dev specs, timelines, social impact — plus an LLM council pattern (subagents or Karpathy clone), synthesis, and Typst PDF outputs for internal and public-facing docs.
Manage one or more Buttondown newsletters from Claude Code — multi-newsletter config, reusable email templates, drafts, sends, subscribers, and API ops grounded in a locally cached copy of the official Buttondown docs.
Choose a professional certification specification-first rather than market-first. create-workspace stands the search up as a private GitHub repo instantiated from the Spec-Led-Certification template, so it survives the machine and its git history shows when the scorecard was frozen relative to when the research ran. Three entry points for the three situations — start-search onboards and runs the full intake, rerun-search archives the previous run and reports what moved in the market, update-profile changes what is stored about you and marks the scorecard stale. Intake writes five dated profile files — subject, current position read from evidence, learning preferences, objectives and standing positions, money and time — a weighted scorecard is derived from those alone and frozen before any credential is looked up, then candidates are scored against it with a source tier and confidence tag on every number. Hard requirements exclude rather than score down. Emits a ranked comparison and a Typst PDF whose figures are computed from the CSVs at compile time. All state is written to the working directory, never to an agent memory store.
ChatGPT Workspace Agents: build, update, inspect, list and share them by driving chatgpt.com/agents/studio in a browser (there is no API for editing agents). Sets instructions, files, skills (one zip per skill; re-upload duplicates, so remove first) and apps. Browser automation: expect breakage when the UI changes; the mechanics doc is dated so drift is easy to fix.
Import a ChatGPT conversation into local files via the browser. Captures the full thread from a chatgpt.com URL, a shared link, or an official export archive, then normalizes it to JSON and renders Markdown or a styled Typst PDF with user and assistant turns marked, artifacts (canvas docs, images, code) extracted, cited sources collected, and conversation metadata recorded. Includes a two-pass redaction skill and a conversation-to-context workflow that writes transcripts into a repo's context/ folder.
Claude-Rudder — collection of utilities to smoothen Claude Code UX. Context-gate workflow, log/blocker capture, plugin/MCP primitives, repo & docs spawning, and the canonical user-data storage convention. (Session-hopping skills moved to Claude-Hopper.)
Claude-Hopper — skills for hopping between discrete terminal-bound Claude Code sessions on Linux. Spawn new instances (Konsole), hand off context (full / clipboard / with-tasks), resume from handovers, and pick up leftover work.
Persona-recipe library for swapping Claude into themed personalities (Daredevil, Jaded IT, Reluctant, Chatty, Philosophical, Operational, Dubious, Hyper Creative, Approval Needed, Visionary, Claude FM, Claude Bouncer). Each recipe ships a banner image and sound effect, and applies via either a managed block in ~/.claude/CLAUDE.md or a repo-sandbox mode that holds the user CLAUDE.md aside.
Safely swap ~/.claude/CLAUDE.md for test/joke configs via symlink. Terminal-only restore that does not depend on the Claude harness, so a hostile test config can never trap you.
Manage the user-level ~/.claude/CLAUDE.md and its chunked ~/.claude/context/ directory — audit, chunk, list, and edit global Claude Code user context for token efficiency.
Software evaluation tables, scoring and quadrants: tick/cross or short-descriptor feature tables, weighted scoring from your stated priorities, and 2x2 quadrant charts, rendered to Markdown, Typst or SVG from one data file and written into the document you're drafting.
Foundational advanced clipboard scripting skills for CopyQ on Ubuntu Linux — CLI reference, custom commands, tab/item management, global shortcuts, and command bundle import/export.
Discover, diagnose, and print to networked CUPS printers from Claude Code. Wraps the lan-mcp-cups MCP server and adds LAN discovery (avahi/Bonjour, arp-scan) plus ufw firewall sanity checks.
First-pass data analysis toolkit: correlations, PII flagging, anomalies, hypothesis tests, data dictionaries, and trend analysis on a dataset in a folder.
Data cleaning, enrichment, restructuring, packaging, and documentation skills for tabular and JSON datasets (no visualisation). 31 skills covering ISO standardisation, PII detection/synthesis, data dictionaries, SQL/graph/vector/HF/GeoJSON/API targets, date & Unicode hygiene, header & numeric-precision standardisation, multilingual header localisation, incremental upstream sync, and Typst-rendered PDF documents of the data.
Streamline programmatic data ingestion against sites and apps that don't ship a documented API — capture network traffic (HAR, mitmproxy, or live tab via claude-in-chrome), map endpoints, infer schemas, and produce a draft OpenAPI spec you can build a stable client against. Ships skills for per-domain map documents, version-controlled storage in a private GitHub repo, and good-faith vulnerability disclosure if a finding turns up incidentally. White-hat use only.
Create static and interactive data visualisations for reports, repos, and data storytelling. Purpose-organised inventory of 60+ validated open-source tools as a head start — static figures (Matplotlib, Seaborn, ggplot2), web charts (Chart.js, ECharts, Plotly.js, ApexCharts, Highcharts), high-perf (uPlot, Perspective, Lightweight Charts), bespoke (D3, Observable Plot, Vega/Vega-Lite, visx, Victory), Python/R apps (Bokeh, Dash, Altair, Streamlit, Gradio, Shiny, D-Tale, Briefer, Preswald), storytelling (Vizzu, VChart, vue-data-ui, SandDance), graphs (G6, sigma.js, Cytoscape, Gephi, Graphviz, GoJS, 3d-force-graph, Constellation), maps (deck.gl, react-map-gl, Leaflet, MapLibre, OpenLayers, folium, react-globe.gl), mobile (fl_chart, F2), BI (Superset, Metabase, Grafana, Kibana, Redash, Chartbrew), diagrams-as-code (Mermaid, PlantUML), domain-specific (Iris, QuantInvestStrats, XCharts, BizCharts, Tablesaw).
End-to-end data annotation toolkit. Prep raw data, design annotation schemas, annotate interactively with Claude (small scale) or scaffold Gemini batch inference (large scale), and publish to Hugging Face.
Apply 20+ classical decision-making frameworks (cost-benefit, pre-mortem, MCDA, decision tree, reversibility, regret minimization, OODA, Eisenhower, SWOT, second-order, opportunity cost, 10/10/10, inversion, base rates, Kepner-Tregoe, six hats, Cynefin, red-team, stakeholder map, time-horizon) to any major decision — parallel multi-lens analysis, synthesis, and Typst PDF export.
Tech diagrams for blogs and docs: Mermaid, text-to-image and image-to-image on fal, Replicate, ComfyUI or your own route, with label and arrow checks.
Start here: browse these plugins by theme, find one for a task, see what's new. Reads the live manifest.
Convert GitHub repos into polished content (PDF, white paper, internal doc, public-docs publication) via Typst, with optional AI banner generation and multi-target publishing.
Companion plugin for the donetick-mcp server. Bundles the MCP and adds skills for chore management against a self-hosted Donetick instance — daily brief, list/create/complete chores, label management.
Uses a real code repository in reverse for developer education: assesses the learner's profile, builds a teaching plan grounded in the repo, writes lessons and file-by-file analyses with code samples drawn from the source, supports interactive Q&A, and typesets any of it as a PDF via Typst.
Design and build responsive HTML email templates using Foundation for Emails, Maizzle, or MJML. Framework-agnostic authoring with email-client compatibility baked in.
Everything right-to-left in one plugin: Hebrew, Arabic, Farsi and Urdu email, Typst PDFs and formal Hebrew letters, HTML marketing newsletters, Google Docs and Sheets, calendar events, Word and Excel files, and render checks. Your signatures, brands and house styles stay in a private user-data folder.
Two-way file exchange between you and Claude: from-human/ and from-agent/ folders, a CLAUDE.md block so Claude uses them unprompted, and a small explorer app (with launcher) locked to that folder, with new-file badges and read receipts.
Plan small custom models and fine-tunes, then save training recipes privately and run them on Modal, RunPod or local Unsloth.
Thin orchestrator for geopolitical forecasting simulations. Scaffolds, runs, bundles, and grades simulations from multiple decoupled upstream templates (lean LLM-council and snowglobe-style actor-simulation variants).
General-purpose digital-evidence processing: capture, hash, OpenTimestamps, ExifTool/MediaInfo metadata, BagIt packaging, immutable sync. Layers with legal-investigative for full chain-of-custody workflows.
Freight volume arithmetic from a product's external dimensions — retrieves and cross-checks dimensions from a supplier listing, then pallet quantity across 11 pallet standards, container loads palletised or floor-loaded for 20 ft / 40 ft / HC / reefer / 13.6 m trailer with stated buffers and payload limits, master-carton permutations ranked by units per container, and freight cost per unit from rates you supply. Reproduces published trade figures; holds no rate data and makes no network call.
Bare-bones GIMP CLI wrapper for Linux: detect install (native/Flatpak/Snap/AppImage), persist a per-user profile, run Script-Fu batch ops, export images, install/list GIMP-side plugins.
Publish content to GitHub gists with clear AI authorship — Claude-authored solution gists and collaborative debug write-ups, both with model+date attribution, environment context, and a PII pre-flight scrub for public publishes. Visibility is a per-call parameter with a configurable default.
Semantic GitHub repo discovery for reusable components. Search, rank, overview, evaluate, and recommend open-source repos — Claude parses gh API results, weighing stars, activity, maintenance, license, and stack fit.
Find, navigate and download Daniel Rosehill's GitHub repositories: search all 1,000+ public repos, find the curated topic index for a subject, and clone one repo or a filtered set. Reads the Start-Here index JSON instead of crawling GitHub.
GPG operations: generate keypairs, export public keys, encrypt, decrypt, sign, and verify files or text using the local GnuPG keyring.
Operational commands and a skill for working with the Green Invoice MCP server — invoices, clients, expenses, and monthly summaries.
Unofficial Guava Health API client: log meds, symptoms and readings, review data, for you and dependents on a family plan. CLI + optional MCP server; DIY NFC tags as a side feature. Not made by Guava.
Habits (rules, always-on or triggered) and SOPs (procedures) as one markdown file each, in a private folder, with one small CLAUDE.md block. Library of example habits to adopt. Replaces sop-library.
Identify and annotate hardware components from photos — circuit boards, motherboards, ICs — with overlays, datasheet cross-checks, and structured BOMs.
Define hardware project BOMs with ESP32-first focus — onboarding captures location/vendors/on-hand gear, then skills for spec creation, live web research of parts (AliExpress/Adafruit/etc.), budgeting, sourcing, compatibility checks, wiring specs, PCB design starting points, assembly instructions, 3D-printable suggestions, and AI-generated mockups via fal.ai nano-banana.
Manage a Home Assistant instance via SSH and the HA REST API — guided first-run onboarding, automation/entity authoring, service calls, TTS testing, and log review. Per-host config is stored outside the plugin so the same install works across multiple Home Assistant environments.
Auto-/clear on a new task: local ONNX classifier moves unrelated prompts to a fresh session. claudehook trial launcher, install, toggle, update-model, remove.
Remove image backgrounds via rembg — single-pass, two-pass cleanup, batch mode, and KDE Dolphin right-click integration.
Capture screenshots and apply annotations (arrows, callouts, boxes, highlights, blur/redaction) on Linux via Pillow + ImageMagick, with batch WebP conversion and PDF bundling. Originals are never modified.
Image production — editing, format conversion, batch ops, and filesystem organisation by resolution, aspect ratio, orientation, format, EXIF time, camera, plus dedupe and metadata scrubbing.
Inventory analysis and decluttering assistant — import a household inventory in any format, then identify discards, duplicates, resale targets, donation targets (geo-aware), insurance-worthy items, and generate throw-out / giveaway lists as CSV or PDF.
Stock counts, aisle/bin and prices across IKEA Israel's 5 branches.
Claude Code agent skills for Israel-specific workflows: emergency readiness utilities, medicine and drug registry lookups, Kol Zchut rights lookups, municipal and transport information, and regional lookups. For Hebrew correspondence and typography see the rtl-skills plugin.
Check opening hours for Israeli businesses, including hours stated relative to Shabbat and yom tov. Combines Google Business and easy.co.il for stated hours with Hebcal candle-lighting/havdalah times, resolving phrasing like 'reopens an hour after Shabbat' into concrete clock times. Bundles 56 Israeli locations with verified per-city candle-lighting customs.
Israel Post (דואר ישראל) appointments, branch search and parcel tracking: book, check or cancel a counter slot.
Israeli shopping workflows — tech retailers (Ivory, KSP, Bug, TMS), Zap price comparison, Hebrew term resolution, ILS conversion, RRP checks, PN cross-reference, brand identification, and AliExpress IL-context search (ILS/Hebrew, IL reviews, free-shipping, combo exclusion, local-vs-import compare).
Maccabi Online (Israel health fund) portal: meds, appointments, doctor messages, results, hitchaivut, refunds. Unofficial; needs your signed-in Chrome.
Cancel Israeli service subscriptions (ניתוק) through netek.co.il — mobile, internet, landline, TV, international calling, water bars, newspapers and credit cards across 55 providers. Resolves the provider and exact Hebrew service string, validates the Israeli ID checksum and address, shows the exact request before sending, and submits only on explicit confirmation. Falls back to filling the form in Chrome when the API changes. Uses Netek's own private, undocumented backend — not a published API.
Look up Jewish texts and references via the Sefaria MCP server — Tanakh, Talmud, Mishnah, Halakha, Kabbalah, commentary, dictionaries, and topics. Includes nikkud add/strip skills (Dicta nakdan + removenikud APIs, offline regex, unikud fallback).
Misc Jewish utility skills: shabbat candle-lighting/havdalah, zmanim (GR"A + MG"A), parsha-of-the-week, Hebrew/Gregorian date conversion (sunset-aware), upcoming holidays (IL/diaspora), and daf yomi. Wraps zmanim-mcp-server and hebcal MCP. Onboarding captures location for halachic-time skills. Companion to jewish-texts-reference.
Give Claude its own KDE Plasma virtual desktop, or a headless sandbox KWin on a private bus: list desktops, create a "Claude" or "AI Agents" desktop, launch and test GUI or terminal apps there without taking over the user's screen (AT-SPI input, KWin window capture), bind the agent desktop to its own display output. Plasma 6 on Wayland.
KDE Plasma (Wayland) runtime utilities — KWin scripting, plasmoid management, panel layout backup, virtual desktops & activities, KGlobalAccel shortcuts, theme/look-and-feel switching, KDE Connect, Klipper, Baloo, kwriteconfig safe-edit, plasma-restart helpers, qdbus introspection, kscreen save/apply, KWallet ops. Complements generic Linux desktop-management plugins.
Skill for developing KDE Plasma plasmoids (QML/Plasma 6 desktop and panel widgets) — scaffold, debug, package, install, and migrate Plasma 5 → 6.
Amazon KDP paperback production with Typst — choose a standard trim, typeset the interior with gutter-aware margins and an even page count, build the full-wrap cover last from the interior's measured extent, pre-flight for silent glyph substitution and stale spines, and assemble the upload folder. Production only; it does not write the book.
Konsole tuned for many agent sessions: window close asks, tab close doesn't; split layout and a Hack-font profile, applied by script.
Print labels from Claude Code on Brother QL and P-touch label printers on Linux. Discovers printers over mDNS, USB and CUPS and keeps a machine-local registry so later prints can name one; renders text and QR labels to the pixel canvas of the loaded DK or TZe media; previews without touching the printer, because brother_ql converts and transmits in a single step with no dry run. Ships a media catalogue of DK and TZe product codes, reusable label templates in the user's data directory, and a Linux driver-install skill covering brother_ql, ptouch-print, mDNS and CUPS. Zebra and DYMO are discovered and registered but not yet printable. Also bundles the original streamable-HTTP MCP server for P-touch setups built around a USB print bridge.
Generates specifications for practice/dummy development projects tailored to the user's learning objectives, technology stack, and proficiency level in each language or tool.
Recommend, generate, and populate software/content licenses. Reads from a user-managed template store and advises on optimal license choice given desired freedoms and constraints.
Keep your computers in step. A git-backed task queue carries installs from one machine to the other, and an SSH comparison finds what has drifted. Successor to os-sync-agent.
Manage Easy Effects on Linux: maintain a preset library, install/export presets, bind autoload to specific mics, test input levels, and set up clean voice-dictation chains. Works with Flatpak or native installs.
Free-key finder + shortcut builder for Ubuntu: probe keys, audit bindings, bind text macros. Checks and installs ydotool/input-remapper, records lsusb IDs, finds non-conflicting keys across KDE, GNOME, input-remapper, keyd and hwdb, and binds by the route that works on X11 or Wayland, then verifies.
Profile Linux keyboards, scan keycodes, and surface underused keys for remapping. Walks intake → keycode dump (xmodmap/XKB/evdev/libinput, X11 + Wayland) → tailored remap suggestions referencing keyd, kmonad, xremap, xmodmap, xbindkeys, input-remapper, autokey.
Linux desktop debugging toolkit — targeted journal/boot/log inspection skills plus proactive logging instrumentation (persistent journald, kdump, sysstat, OOM protection) so AI agents can analyze hard crashes, freezes, and runtime issues. Targets Ubuntu + Wayland; forkable for other distros.
Linux packaging and release workflows — Debian/.deb builds, npm publishing, GitHub release creation, agent deploy scripts, and local debugging
Performance and space optimisation for Linux desktops — hardware-aware CPU/GPU/disk/memory benchmarks with governor / I/O scheduler / sysctl tuning, plus disk-usage analysis (BTRFS-aware), duplicate-file detection, package audit, and dev-clutter pruning (venvs, node_modules, caches).
Batch inference, cloud or local: Ollama and vLLM backends, batch vs sync cost, OpenRouter model lists, model evals, and runners with spend caps.
Scaffold new LLM Council projects from existing templates (Template, Grounded, Decide) or build bespoke council repos for specific purposes.
Interface with a Music Assistant server via its local API — onboard a deployment, control players, snapshot speaker rosters, save/update per-player DSP presets, and apply a curated podcast EQ preset.
Trace an IoT device's cloud egress, document it, lock it down: IP cameras, OPNsense, packet capture, whois, block rules, verify.
NFC tag operations using libnfc — read, write, inspect, password-protect, and bulk-write from CSV with manual tag-by-tag feed.
Skills for analyzing large text corpora — topic modeling (BERTopic with temporal evolution), NER, categorization into fixed taxonomies, bottom-up category derivation, multi-level taxonomy design, word frequency, synonym clustering for voice-note/STT corpora, parametric stats, and metadata↔content correlation. Three execution lanes (classical NLP, local LLM via Ollama, cloud LLM via OpenRouter) with explicit cost-awareness: mandatory pre-run estimates for >1k-doc LLM passes, two-pass cheap→premium pattern, embeddings+clustering preferred over pairwise LLM comparison.
Read OBD-II data from an ELM327-class adapter, normalise it to JSON, and use it to diagnose faults and plan vehicle maintenance. WIP.
Manage OBS Studio on Linux from Claude Code: detect install type (native/Flatpak/Snap/AppImage), enable and validate obs-websocket, ship a bundled obs-mcp MCP server for programmatic OBS control, back up configs to a user-defined folder, and install third-party OBS plugins.
Research, discover, compare, and evaluate AI models on OpenRouter — backed by the bundled Model-Scout MCP server for live catalog data with caching. Subsumes the standalone open-router-model-research plugin and the Model-Scout-MCP server. 11 skills cover lookup, capability filtering (tools, vision, audio), recommendation, head-to-head comparison, deep evaluation, workload cost projection, and finding cheaper alternatives.
Manage an OPNsense router/firewall via SSH and the OPNsense API — guided first-run onboarding, firewall rule inspection, network debugging, and host/log diagnostics. Per-host config is stored outside the plugin so the same install works across multiple environments.
Plan and burn batch M-Disc / BD-R / DVD archives from a source directory on Ubuntu, with copy multipliers for offsite duplicates. Wraps growisofs, xorriso, and dvd+rw-mediainfo; optional K3B handoff for manual fallback.
Evaluate AI models for a real job, starting from start-eval: define the objective and workload, draw candidates from the live OpenRouter catalogue, and run wide and shallow (many models, about two outputs each). Covers in-repo and agent-run eval spaces, framework CLIs, image, RAG, agent, speech and regression evals, private annotation and human feedback Spaces, and findings through to articles.
Browser-driven AliExpress shopping assistant for Claude in Chrome. Bundles userscripts (Find Similar, Hide Combo Deals) and skills/commands that drive aliexpress.com directly in the user's browser session — search, read current listing, compare open tabs, apply filters, hide bundles, and find similar items. Scope is global aliexpress.com (any locale/currency) and visual/interactive shopping, not headless scraping or programmatic API flows.
Anonymise sensitive documents (whistleblower disclosures, harassment diaries, incident reports) by replacing identifying details while preserving narrative integrity.
Speaker diarization, voice enrollment, and identification for audio files — builds a local voiceprint library with optional cloud replication.
When you hit a wall, spin up a research workspace to brainstorm solutions — captures the blocker, what you tried, and kicks off deep research
Conda environments — list, validate, compare, back up, audit and clean up environments with AI assistance.
A workflow planning and time management template with agent definitions and slash commands for managing daily, weekly, and monthly schedules while maintaining connections between immediate tasks and longer-term objectives.
Slash commands and subagents for managing Docker containers, Compose stacks, volumes, networks, and multi-environment deployments.
Process .eml email exports — reconstruct a thread from loose .eml files, render the chain to a single PDF with correct right-to-left handling for Hebrew and Arabic, extract attachments, export Markdown, redact before sharing, and build numbered evidence bundles.
File system organisation — folder structure optimisation and file management automation.
Research existing GitHub repositories before building — search, rank, and evaluate candidate tools via gh CLI with careful attention to stars, recency, and maintenance quality
Grainger.com (US industrial MRO) product research — typeahead and category discovery from a script, price and spec verification from the signed-in browser. Encodes the mapped /ta/v3, /tap/v2 and PRELOADED_STATE surfaces, the facet grammar that filters a category, per-branch stock counts across ~325 branches, and the three distinct ways Grainger refuses a request while returning HTTP 200.
Learning resources, skill development, educational content creation and knowledge management.
Observe and interact with an MQTT broker — monitor all payloads, watch a specific topic, and publish messages. Stores broker credentials once and reuses them.
Spawn a new stack search workspace from the Claude-Stack-Research-Workspace template — clone, name, create GitHub repo, and kick off the structured intake interview
Technology research — competitive analysis, technical documentation retrieval and research workflows.
Personal user manuals and private documentation for codebases, with PDF output.
Skills and an orchestrator agent for preparing PDFs for digital printing — resize, grayscale, font embedding, transparency flattening, image downsampling, color normalization, watermarks, footer burn-ins, cover pages, bleed-safety check, job folders, formal print orders, and email/Drive share.
Convert PDFs to clean Markdown, chunk into logical sections (chapters, indexes, appendices), and extract embedded tables to CSV. Local-first via marker/docling/pymupdf4llm + camelot/tabula, with TOON manifests.
Claude Code skills for taming Pipewire/Wireplumber audio on Linux — manage default devices, persistent device-priority rules, per-app routing, mic level checks, and EasyEffects bindings.
Nudges Claude to document decisions, todos, sprints, and handovers into a structured planning/ tree by default — without being asked each turn. Ships a home-level CLAUDE.md snippet plus /nudge-install and /nudge-scaffold slash commands.
Keep a repo's decision log and plans lean: record numbered decisions, and compact old decisions and plans into in-place summaries so agents read the live state, not the history.
Site skills built by watching a site in your own signed-in Chrome — network-first discovery, a surface survey or your own URL list, a skill roster for approval, then skills that survive UI changes, keep the session alive, clean up after themselves and probe for drift. Semantic handles only. Every observation is classified publishable, private, or never-recorded.
Curated batches of third-party Claude Code plugins, grouped by type/theme, installable in one command.
Scan files, directories, or git repositories for personally identifiable information — credentials (gitleaks), generic PII (Microsoft Presidio), and matches against a user-maintained personal PII inventory (names, addresses, family, IDs) stored locally.
Personal schedule, task, and meeting management. Routes mixed brain-dumps into Google Calendar (events) and Todoist (tasks); manages agenda/minutes Google Docs linked to events; produces wrapup logs and morning briefs. 22 skills covering Calendar/Todoist CRUD, firehose routing, task<->event migration, priority/date hygiene, and agenda/minutes lifecycle.
Jump to Daniel Rosehill's public work from Claude Code — profiles (LinkedIn, X, GitHub, YouTube, npm, Kaggle, DeviantArt, Pexels), blog, Hugging Face, public repos by recency/stars/A–Z, bio, resume download, and contact/newsletter pages. Also refreshes this marketplace's local cache. A directory of one person's published output, not personal configuration.
Capture mid-task ideas, bugs and asks into sprints, keep going. Triages each item (category, priority P0-P3) into the current sprint, next sprint, backlog or ideas in an in-repo planning/ scaffold, then resumes the task in hand. Also plans sprints, gates releases until the sprint is done, and answers what's next.
Manage a Proxmox VE host via SSH and the Proxmox API — guided first-run onboarding, VM/CT lifecycle, storage and ZFS inspection, log review, and update workflows. Per-host config is stored outside the plugin so the same install works across multiple Proxmox environments.
Routine rclone admin: install/update, add/list/edit/delete remotes, add/list/edit/remove mounts as systemd user units, config backup (one-off or automatic to a private git repo on a timer), migrate config between machines with path rewriting, and docs lookup.
AI-assisted Linux system recovery — slash commands and agents (diagnose, logs, network, disk, services, packages) for diagnosing and fixing a broken system. Pairs with an optional GRUB/systemd installer that boots a minimal recovery TTY straight into Claude CLI.
Manage a resume as JSON Resume schema data and render it with custom Typst templates. Onboard, iterate, fork variants, and version.
Capture geo-restricted web content (articles, prices) via the user's own localhost so requests exit via the user's IP. Headless-first escalation ladder (Scrapling static -> stealth -> Playwright -> real Chrome via bb-browser). Project-local save (in-repo captures/) with global fallback. Batch capture with human + agent summaries, Typst PDF compilation, and arbitrary-language capture+translate (default Hebrew -> English).
Terminal crashed? Reopen the lost Claude Code sessions: finds the ones that died together and resumes each in its repo, in one Konsole window of split panes (or tabs, or windows). Also keeps Konsole asking before a window close kills them.
Ask 'should this become a skill?' before every task: a five-step checklist (repeatable, automatable, already ours, already someone else's, else create and validate it) plus an installer that puts the heuristic in your user-level CLAUDE.md / AGENTS.md.
Project-scoped authoring tools for Standard Operating Procedures and decision flowcharts. Scaffold from templates, embed Mermaid/D2 diagrams, compile to printable PDFs via Typst, and assemble multi-document binders with TOC and page numbers.
Spec-driven development workflow for Claude Code: turn unstructured project briefs (especially voice transcripts) into a versioned spec, modular context, and a CLAUDE.md scaffolded into your current repo.
Component-level review of a project's technology stack, not its code — records the stack to docs/stack/ with the run date, checks each component for end-of-life and currency, finds architectural gaps (caching, backups, queues, rate limiting, observability), flags redundancy and over-engineering against the real workload, then emits a sequenced revision proposal and an execution brief. Re-runnable per repo with persistent decisions and finding history.
Loose skills that will be migrated into other plugins later. Recommended not to enable/use this!
Contact a vendor's support desk from a registry of verified, dated contact routes rather than recalled ones — ticket URLs and which support plans can reach them, abuse and fraud addresses taken from mailto hrefs rather than page text, per-tier response windows, priority vocabulary, escalation ladders with SLA claim deadlines, pre-contact checks, and the plausible routes that turn out to be dead. Records the user's account identity and where credentials live (a pointer, never the secret), drafts the request with the identifiers the desk will demand, sends it through whichever channel is actually reachable, logs the ticket and chases it when the window elapses. Ships verified for Twilio only; extend with support-add-service. Work in progress — concept stage.
Manage a Synology NAS via SSH — guided first-run onboarding, share/volume inspection, storage health, and file operations. Per-host config is stored outside the plugin so the same install works across multiple NAS environments.
Generate synthetic datasets from schemas, real data, or LLM-driven personas. Tabular fit-and-sample (SDV: GaussianCopula, CTGAN, TVAE), Faker/Mimesis schema generation, deterministic PII swap, LLM-driven real-to-synth conversion for unstructured records, and SDMetrics-based quality/privacy evaluation (plus embedding-based leakage checks for text).
Repo-based task queueing system with categorisation and prioritisation. Scaffolds a planning/ folder, logs tasks (single, batch list, or voice-transcribed), buckets them by category, and hands prioritised work off to the repo's orchestration agent.
Generate taxonomy and lookup tables (countries, currencies, languages, US states, timezones, custom domain taxonomies) and load them into Postgres, SQLite, MySQL, or export to CSV/JSON/SQL seed files. For data engineers, CMS builders, and eval pipeline authors.
Rewrite text in deliberately ridiculous styles — Shakespearean, medieval, archaic, chaos-case, over-salesy, platitude-stuffed, pseudobot, plus length transforms (elongate / truncate). Nine no-config skills for stylistic mischief.
Timemark stamp-camera photo import that recovers the field data the app withholds. No Timemark export carries both the full-resolution image and the values, so imports are gated on having the originals and the Photosheet XLSX, then joined on capture timestamp. Outputs are XMP burned into the files or a self-contained archive of images, JSON sidecars and an index.
Overflow /tmp to a capped, self-wiping NFS share on a home server. For laptops whose /tmp is a small tmpfs (RAM) or whose disk is tight: an ext4 image exported over NFS, mounted on demand with a local-disk fallback, expired hourly after 24 h. Skills to set it up, route large temporary files there, and report and clean it.
Convert JSON, CSV, YAML, and other structured data into TOON (Token-Oriented Object Notation) — a compact, lossless re-encoding that uses ~40% fewer tokens than JSON when fed to LLMs. Wraps @toon-format/toon and tracks the wider ecosystem (Python, Java, .NET, PHP, Rust ports).
Make optimal use of a baggage allowance on an international itinerary — research each carrier's real checked, cabin and personal-item allowance per segment with a source and a confidence level for every figure, compute the binding limit that governs the journey, inventory what is carried by weight, reconcile scale readings against the inventory, solve the allocation to bags and packing cubes, price every way out of being over, and rank an emergency leave-behind list by cost per kilogram saved. Renders a one-page airport counter card and a per-bag packing list as PDF via Typst and as Markdown. One trip workspace per itinerary.
Local text-to-speech: Kokoro, NeuTTS Air, Chatterbox. Installs to fit your hardware (CPU or ROCm/CUDA GPU), one-shot voice cloning with stored embeddings, read-aloud and audio files.
Skills for preprocessing text for TTS engines — SSML conversion, ElevenLabs markup (grounded in live ElevenLabs prompting docs), TTS safety review, and manual prosody addition. Non-destructive by default: edits land in an edited/ folder alongside an unchanged original/ copy.
Build programmatic document generation pipelines — Typst templates for local batch rendering, plus integration scaffolding for n8n and cloud rendering services (Carbone, PDFMonkey, APITemplate, DocRaptor, Docmosis, Adobe Doc Gen).
Typst PDFs: docs, briefings, screenplays + your own private formats. Built-in templates (personal, public, business, client-confidential, briefing sheet, spec screenplay) with a private overlay folder for your identity, logo and custom formats, so you extend it without forking.
Administer Ubuntu machines, desktop or server, locally or over SSH: health and diagnosis, packages (apt/snap/flatpak), systemd services and logs, Docker, storage and backups, networking, hardware and GPU drivers, displays, fonts, codecs, shell and dev environments. Supersedes linux-desktop-mgmt, linux-server-mgmt and sysadmin-homelab.
Host security for Ubuntu desktops and servers: ufw firewall (setup, status, scoped allows, removal, diagnosing blocks), ClamAV/rkhunter/Lynis/AIDE scanning and schedules, security audits and updates, SSH keys and sshd config, fail2ban, YubiKey and fingerprint login, TLS certificate checks. Supersedes linux-av-manager and security-checkup.
Product manuals cut to what you need: your language, no warranty boilerplate, or a one-task how-to PDF. Fetches the official PDF, trims pages, rewrites via OCR, and cites the original in every output.
Develop, test, and publish Tampermonkey userscripts — scaffolding, in-browser validation via Claude in Chrome, README generation, and version bumping.
Report bugs, request features, disclose security to vendors: per-vendor channel data with routing rules, a private overlay for your preferred channels and personal contacts, duplicate checks before filing. Vendors: Claude Code, 1Password, Whisperian.
Plan and prompt-engineer AI-generated visuals (images, diagrams, video) for whitepapers, blog posts, and long-form content. Six skills cover project onboarding, visual ideation, prompt generation, project listing, fal-ai execution, and a resolution/style reference.
Passive-aggressive PA system. Claude barks status updates over a speaker; if ignored, escalates across the house via Home Assistant or MQTT. Includes pre-recorded voice packs, RGB signal bulb, full-screen flash overlay, and a quiet-mode skill that translates natural-language pause/schedule requests.
Voicenotes.com: search, read, create; unofficial text and date-range export, tags, audio import. Not affiliated.
Maintain a home Zigbee network — onboard MQTT broker, coordinator (SMLight / Sonoff / ConBee / etc.), and Home Assistant; manage credentials, network exports, and routine maintenance.
Scaffolds used to be distributed as ~100 standalone template repos and a New-Repo-From-Template plugin. That pattern was retired in the April 2026 reshape.
Scaffolds now live inside the cluster plugins. Each of the 28 cluster plugins bundles the workspace primitives for its domain (commands, skills, agents, MCP configs) and exposes a provisioning skill that writes a fresh per-project scaffold on demand — so instead of cloning a template repo, you install the relevant plugin once and ask Claude Code to scaffold a new workspace for whatever project you're starting.
See the Plugins section above for the full cluster list.
Individual slash commands, sometimes integrated into other plugins or sometimes just for use at the user level.
See also: The Claude Slash Commands repo serves as both a 350+ command library and the dedicated index for all slash command repos.
Adds AI/human attribution sections to README files for transparent tool usage documentation.
Slash command deploying a sub-agent for codebase optimization and dead code removal.
Flips Claude Code into autonomous, action-first mode — ships a per-session /calls-the-shots slash command plus an optional always-on snippet injected into ~/.claude/CLAUDE.md.
System administration slash commands for Linux desktop environments.
Slash command that condenses bloated CLAUDE.md files to essentials and organizes supplementary context into a structured agent-context/ folder.
General-purpose slash command library for various Claude Code workflows.
Drop-in system prompt and slash command that make coding agents capture what they discover — undocumented APIs, auth flows, dead ends — instead of losing it when the session ends.
Early standalone Claude Code plugin with slash commands and agents for code generation, refactoring and general development.
Early standalone Claude Code plugin with slash commands and agents for Git version control and GitHub repository management.
Slash commands encouraging use of existing libraries instead of building custom solutions.
Slash command that prunes a repo's CLAUDE.md to essentials and scaffolds a topic-based context/ store, an ingestion pipeline for raw materials, and subagents for context maintenance.
Other projects including meta-resources, feedback, and utilities that span multiple categories.
Index of repositories that each record one agentic workflow being run end to end on a real task, including the wrong turns, as reference material for designing plugins.
Public archive of structured AI ideation sessions run via Claude Code — each /run takes a topic and saves a dated batch of ideas with name, summary, category, feasibility/impact ratings, considerations, and next steps.
Planning notes and architecture diagrams exploring three complementary methods for building a personal AI memory system over a vector store — passive distillation from chat, agentic interviews, and manual curation.
Staging repo for AliExpress skills (Israel buyer focus) before they ship in the Israel Shopping plugin: scraper changes, locale handling and Affiliate API signing.
Claude Code plugin for browser-driven AliExpress shopping via Claude in Chrome — search, read listings, compare tabs, filter, and find similar, plus bundled userscripts.
Workspace tracking a blog post through iterative drafting and revision, keeping every iteration, diff, and supporting context.
Claude Code plugin for moving or copying media off a connected Android phone using user-defined mappings, MIME-based routing, and optional date-folder and orientation-split layouts.
Staging area for Claude Code bug reports and feature requests before filing upstream at anthropics/claude-code.
Ad-hoc close readings of selected entries from the upstream Claude Code changelog, with generated PDFs and categorised JSON per version.
Small tricks and temporary workarounds for Claude Code, each timestamped because the harness usually closes the gap quickly.
Running notebook of techniques for getting more out of Claude Code.
Meta-plugin for managing Claude Code configuration, slash commands, and agent development.
Centralized marketplace for discovering and installing Claude Code plugins — source of truth for the Plugins section.
Feature requests for improving Claude Code's context handling capabilities.
Dated public notebook of friction points encountered across Claude Code, the desktop and web apps, and the API, with point-in-time documentation snapshots used as grounding.
Transcript from an Anthropic user research interview on AI tools and adoption.
Template for creating curated resource lists with automated formatting and badge generation.
Index of people and projects building with Claude, including Claude Code, in Israel.
Agent-readable search manifests documenting advanced URL parameters for websites, written for AI agents rather than human readers.
System prompt and skill notes for keeping repository boundaries clean during agentic development.
Pattern for AI-assisted development that starts from a single long user prompt and refines it into a spec.
Plugin for generating personalized user manuals and private documentation for codebases.
Notes on the Claude Code voice mode feature.
Plan and records from the April 2026 refactor that turned roughly 100 standalone Claude workspace template repos into cluster plugins: inventory, clusters, conversion log and deletion list.
Claude Code-powered maintenance tool, with agents and slash commands, for a hierarchy of GitHub index repos (master index, index of indices and subindices).
Index of the Israel agent-skills ecosystem: released Claude Code plugins, the work-in-progress skills workspaces that feed them, and related indexes.
Claude Code plugin for recording Israeli phonebook entries in Google Contacts — counter versus call-centre hours, Shabbat and chag closures, seasonal timetables, and numbers that do not connect from abroad.
Research and notes on installing Claude Code plugins locally rather than through the marketplace.
Tool that determines whether to continue an AI conversation or start fresh for optimal context.
Showcase of creative Claude Code applications beyond traditional software development.
Draft code of practice for publishing agent skills that touch the open web, separating the runtime information a skill needs from the research an author gathered about a site's defences.
Research and notes on designing and authoring Claude Code skills.
Early-stage repositories: ideas, planning notes, placeholders and work-in-progress skills baskets that have not yet become a working tool, workspace or plugin. Some are stubs. They are listed so the idea is findable, not because there is anything to install.
Idea: a per-turn router that sends each message to the current task, an existing parallel worker or a new session, keeping unrelated tasks out of one agent's context.
Placeholder for planning a Claude Code background hook.
Empty placeholder for experiments with CLAUDE.md structure.
Stub for planning notes on the plugins and skills mechanism (May 2026).
Planning repo for Claude placing a real outbound phone call to deliver a one-time, task-specific spoken message on the user's behalf.
Stub workspace for agent skills that check medication availability in Israel for patients, to be folded into the Israel agent-skills plugin.
Planned multi-label classifier that routes voice-note transcripts to task-specific subagents (email, calendar, shopping list, to-do, dev spec), each holding one MCP tool.
