AI Agent Skills

An opinionated selection of skills for daily dev workflows.

Agent Skills Compatible - All skills follow the Agent Skills open format specification.

Available Skills

Skill	Description
umbrel-app	Expert assistant for developing, packaging, testing, and submitting apps for umbrelOS
claude-code-expert	Comprehensive Claude Code & Anthropic ecosystem knowledge. Official patterns for agents, skills, hooks, commands, MCP.
agent-browser	Expert on agent-browser - Vercel's headless browser automation CLI for AI agents with 50+ commands, snapshots, and multi-session support
chat-sdk	Expert on Chat SDK (vercel/chat) - the unified TypeScript SDK for building chat bots and agents across Slack, Teams, Discord, Telegram, WhatsApp, and more
ai-sdk	Expert on AI SDK - Vercel's TypeScript toolkit for building AI applications with unified LLM API, streaming, tool calling, and agents
agent-skills	Expert at the Agent Skills open format for extending AI agent capabilities - create, validate, and understand SKILL.md files
hermes	Expert at understanding and working with Hermes Agent - its memory system, skills, cron jobs, tools, and behavioral conventions
honcho	Expert on Honcho — AI-native memory and context platform for LLM applications with persistent memory, user modeling, and context management
safe-delete	Prevents catastrophic file deletion by transforming rm commands to trash and blocking dangerous patterns like `rm -rf /`
x-engagement	Crafts high-engagement X (Twitter) content using the published X algorithm scoring weights, visibility filtering and account-standing rules, conversation tactics, and Original Content Rewards monetization
gh-issue-tracker	Install, configure, and manage gh-issue-tracker — lightweight error tracking that creates GitHub Issues with deduplication, fingerprinting, and rate limiting
firecrawl	Expert on Firecrawl — web scraping, crawling, search, and browser automation API for AI agents with clean LLM-ready output
servarr	Deploy, configure, and manage the full media stack — Sonarr, Radarr, Lidarr, Prowlarr, Plex, Overseerr, qBittorrent, Bazarr, and Recyclarr
obsidian	Manage and optimize Obsidian vaults — organization, Dataview, Templater, workflows, MCP integration, plugins, sync, and publishing
adguard	Deploy, configure, and manage AdGuard Home — network-wide DNS ad blocking, filtering, DHCP, client management, and REST API automation
immich	Deploy, configure, and manage Immich — self-hosted photo and video management with machine learning, facial recognition, mobile backup, and REST API
glinet	Configure and manage GL.iNet routers — VPN, AdGuard Home, DNS, multi-WAN failover, drop-in gateway, firewall, and network modes
umami	Deploy, configure, and manage Umami — open-source privacy-focused web analytics with API client, tracker, events, statistics, and reports
flue	Expert on Flue — the open-source TypeScript framework for building durable, autonomous AI agents and workflows on the Pi harness. Write once, deploy anywhere, use any LLM.
pi	Expert on Pi (pi.dev, @earendil-works/pi-coding-agent) — extend, improve, custom providers, extensions, skills, themes, and running from source
wealthfolio	Expert on Wealthfolio — the open-source, private, local-first portfolio & net-worth tracker (desktop, iOS, self-hosted Docker). Concepts, CSV import, self-hosting, and addon development.
1password-cli	Expert on the 1Password CLI (`op`) — manage 1Password from the terminal, read/inject secrets with secret references, run op run/op inject, manage items & vaults, service accounts, shell plugins, and the SSH agent.
portainerctl	Expert on portainerctl — Portainer's official CLI for driving Portainer Business Edition over its REST API. Auth via API token, environments, stacks, GitOps deploys, containers, Kubernetes, edge, users/teams/RBAC.
beads	Expert on beads (`bd`) — the dependency-aware, Dolt-backed issue tracker built for AI coding agents. Ready-queue workflow, hash IDs, dependency types, formulas/molecules/gates/wisps, Dolt sync, and Claude Code wiring
tower	Run a control tower over a fleet of herdr agents — session start, catalog, unpaid ask, dispatch from a spec file, watch a completion marker, reconvene with an owner table, and verify through acceptance gates so "done" is an exit code
pr-standard	House standard for GitHub pull request descriptions — the six ASD-STE100 writing rules, the four-label pack (type, `priority:`, `t-shirt:`, `area:*`), the required body shape with a mermaid diagram, and the forked-chat workflow
skill-release-preflight	Three checks before pushing to this repo — never hand-edit a release-please version, run `check-consistency.sh` locally, and read `gh label list` before `gh pr create`
apprise-api	Deploy, configure, and drive the Apprise API (caronc/apprise-api) — the self-hosted notification gateway that fans one HTTP request out to 100+ services, with stateful config keys, tag routing, attachments, and webhook payload mapping

Installation

Recommended: install with zskills (Rust CLI package manager)

zskills is a Rust package manager built specifically for managing Claude Code skills — declarative install, multi-marketplace, atomic settings.json round-trip.

cargo install --git https://github.com/zot24/zskills

zskills marketplace add zot24/skills
zskills install umbrel-app servarr immich    # name resolution is unambiguous

Declarative skills.toml in any repo or in ~/.config/zskills/:

[[skills]]
name = "umbrel-app"
marketplace = "zot24-skills"

[[skills]]
name = "servarr"
marketplace = "zot24-skills"

zskills sync                  # idempotent, applies the manifest
zskills sync --dry-run        # preview changes first

Inside Claude Code (built-in)

/plugin marketplace add zot24/skills
/plugin install umbrel-app@zot24-skills

Or add to project settings manually

.claude/settings.json:

{
  "extraKnownMarketplaces": {
    "zot24-skills": {
      "source": { "source": "github", "repo": "zot24/skills" }
    }
  },
  "enabledPlugins": {
    "umbrel-app@zot24-skills": true
  }
}

Usage

Once installed, you can use skills via slash commands or natural language.

Slash Commands

Format: /skill-name:command-name [arguments]

# Umbrel app development
/umbrel-app:umbrel-app scaffold my-app               # Create new app structure
/umbrel-app:umbrel-app validate ./my-app             # Validate app configuration
/umbrel-app:umbrel-app convert ./docker-app          # Convert Docker Compose to Umbrel
/umbrel-app:umbrel-app pr ./my-app                   # Generate PR submission
/umbrel-app:umbrel-app debug ./my-app                # Troubleshoot issues

# Claude Code expertise
/claude-code-expert:claude-code-expert create agent  # Guide for creating agents
/claude-code-expert:claude-code-expert create skill  # Guide for creating skills
/claude-code-expert:claude-code-expert validate ./x  # Validate against best practices
/claude-code-expert:claude-code-expert features      # Show Claude Code capabilities

# Agent browser automation
/agent-browser:agent-browser open <url>              # Open a webpage
/agent-browser:agent-browser snapshot      # Get element refs
/agent-browser:agent-browser click @e2     # Click by ref
/agent-browser:agent-browser screenshot    # Capture viewport

# Hermes Agent
/hermes:hermes memory                    # How Hermes memory works
/hermes:hermes skills                    # How to create Hermes skills
/hermes:hermes cron                      # Cron job configuration

# Honcho memory platform
/honcho:honcho quickstart                # Get started with Honcho
/honcho:honcho context                   # Context retrieval for LLMs
/honcho:honcho sdk                       # SDK reference

# Safe delete
/safe-delete:safe-delete enable          # Enable trash-based deletion
/safe-delete:safe-delete status          # Check current mode

# Firecrawl web scraping
/firecrawl:firecrawl scrape https://example.com  # Scrape a page
/firecrawl:firecrawl search "AI news"            # Web search
/firecrawl:firecrawl crawl https://docs.site.com # Crawl a site
/firecrawl:firecrawl mcp                         # MCP server setup

# Media stack management
/servarr:servarr setup docker      # Full Docker stack
/servarr:servarr integrate         # Wire up all apps
/servarr:servarr profiles          # TRaSH quality profiles
/servarr:servarr remote            # Laptop→NAS setup

# Immich photo management
/immich:immich setup                  # Docker Compose deployment
/immich:immich backup                 # Database + filesystem backup
/immich:immich library /mnt/photos    # External library setup
/immich:immich upload /path/to/photos # CLI bulk upload

# Apprise notification gateway
/apprise-api:apprise-api deploy           # Docker / Compose / Kubernetes setup
/apprise-api:apprise-api configure        # APPRISE_* environment variables
/apprise-api:apprise-api key my-alerts    # Manage a stateful config key
/apprise-api:apprise-api troubleshoot 424 # Diagnose a response code

Natural Language

You can also just describe what you want:

"Create an Umbrel app for my Docker project"
"Help me package this for umbrelOS"
"How do I create a Claude Code agent?"
"What are the best practices for hooks?"
"Automate browser login with agent-browser"
"How do I use snapshots for element selection?"
"Add persistent memory to my LLM agent with Honcho"
"How do I get user context from Honcho for my chatbot?"
"Write a viral X/Twitter post about AI"
"Set up a safe delete alias so I don't accidentally delete everything"
"Deploy an Apprise API container and route my alerts to Discord and ntfy"

The skills auto-activate based on context.

Skill Details

umbrel-app

Full lifecycle support for Umbrel app development:

scaffold - Generate complete app structure
validate - Check against 20+ requirements
convert - Transform Docker Compose to Umbrel format
pr - Generate submission-ready PR content
debug - Troubleshoot issues
sync/diff - Stay updated with upstream docs

/umbrel-app:umbrel-app scaffold my-app
/umbrel-app:umbrel-app validate ./my-app

Full documentation

claude-code-expert

Comprehensive Claude Code and Anthropic ecosystem knowledge:

create agent/skill/hook/command - Official patterns and guidance
validate - Check artifacts against official checklists
patterns - Browse available creation patterns
features - Learn about Claude Code capabilities
sync/check - Keep documentation up to date

/claude-code-expert:claude-code-expert create agent
/claude-code-expert:claude-code-expert create skill
/claude-code-expert:claude-code-expert validate ./my-agent
/claude-code-expert:claude-code-expert features

Full documentation

agent-browser

Expert on Vercel's browser automation CLI for AI agents:

open - Navigate to webpages
snapshot - Get accessibility tree with element refs
click/fill/type - Interact with elements by ref
screenshot - Capture viewport or full page
session - Manage isolated browser instances
network - Intercept and mock requests
sync/diff - Stay updated with upstream docs

/agent-browser:agent-browser open https://example.com
/agent-browser:agent-browser snapshot -i
/agent-browser:agent-browser click @e2
/agent-browser:agent-browser screenshot page.png

Full documentation

chat-sdk

Expert on Chat SDK (vercel/chat), the unified TypeScript SDK for multi-platform chat bots:

overview - Introduction, core concepts, and getting started
adapters - Platform adapters for Slack, Teams, Discord, Telegram, WhatsApp, and more
ai - AI streaming, AI SDK tools, and message conversion
api - Look up Chat/Thread/Channel/Message API reference
sync/diff - Stay updated with upstream docs

Full documentation

ai-sdk

Expert on Vercel's AI SDK - TypeScript toolkit for building AI applications:

providers - Unified API across 100+ LLMs
streaming - Streaming text, audio, and image generation
tool calling - Structured tool definitions and execution
agents - Build autonomous agents with memory
rags - Retrieval augmented generation patterns
sync/diff - Stay updated with upstream docs

Full documentation

agent-skills

Expert at the Agent Skills open format specification:

create - Create new SKILL.md files following the spec
validate - Validate skills against the specification
spec - Understand the Agent Skills format
best-practices - Authoring guidelines and patterns
integrate - Add skills to your agent
sync/diff - Stay updated with upstream docs

/skills-ref validate ./skills/my-skill

Full documentation

hermes

Expert at understanding and configuring Hermes Agent:

memory - Dual-store memory, Honcho profiles, instruction capture
skills - How skills are created, triggered, and structured
cron - Schedule formats, delivery targets, scheduler architecture
tools - Built-in tool categories and capabilities
config - Configuration files and environment variables
conventions - Communication style and user preferences

/hermes:hermes memory
/hermes:hermes skills
/hermes:hermes cron

Full documentation

honcho

Expert on Honcho — AI-native memory and context platform for LLM applications:

quickstart - Setup and first API calls
architecture - Core data model (workspaces, peers, sessions)
context - Retrieve context for LLM injection
peers - Working with peer representations
dreaming - Autonomous memory consolidation
sdk - Python and TypeScript SDK reference
api - Full REST API reference
integrate - Guides for Claude Code, MCP, LangGraph, Discord, Telegram, and more
sync/diff - Stay updated with upstream docs

/honcho:honcho quickstart
/honcho:honcho context
/honcho:honcho sdk
/honcho:honcho integrate discord

Full documentation

safe-delete

Prevents catastrophic file deletion by intercepting dangerous rm commands:

enable - Activate trash-based deletion (aliases rm to trash-cli)
block - Block patterns like rm -rf / or rm -rf /*
whitelist - Allow specific directories to be deleted
status - Show current protection state
restore - Recover files from trash

/safe-delete:safe-delete enable
/safe-delete:safe-delete status

Full documentation

x-engagement

Crafts high-engagement X (Twitter) content using strategic frameworks:

hook - Write viral-worthy opening hooks
thread - Draft a full tweet/thread
reply - Craft strategic, Grok-quality-aware replies
cluster - Plan a content cluster of related posts
review - Review a draft for negative-signal risk, action triggers, and slop
ideas - Generate ideas ranked by algorithm signal priority
signals - Explain the published scoring weights and how to read them
weights - Print the published blend weights with file:line citations
standing - Diagnose a reach drop via visibility filtering and account labels
monetization - Original Content Rewards eligibility, qualified impressions, payout rules

/x-engagement:x-engagement hook "AI agents"
/x-engagement:x-engagement thread "Why most AI projects fail"
/x-engagement:x-engagement monetization

Full documentation

firecrawl

Expert on Firecrawl web scraping, crawling, and data extraction API:

scrape - Extract markdown/HTML/JSON from a single URL
search - Query-based web discovery with content extraction
crawl - Recursively gather content from entire sites
map - Discover all URLs on a website
interact - Browser automation (clicks, forms, navigation)
extract - LLM-powered structured data extraction
sdk - Node, Python, Go, Rust SDK reference
mcp - MCP server setup for Claude/Cursor
sync/diff - Stay updated with upstream docs

/firecrawl:firecrawl scrape https://example.com
/firecrawl:firecrawl search "AI news"
/firecrawl:firecrawl crawl https://docs.example.com

Full documentation

servarr

Deploy and manage the complete media automation stack:

setup - Docker Compose for full stack (9 services)
configure - App-specific configuration guides
api - API usage with curl examples for all apps
integrate - Wire up Prowlarr → *arr apps → qBittorrent → Plex
profiles - TRaSH quality profiles via Recyclarr
status - Health check all running instances
troubleshoot - Diagnose common issues
add - Add movies/shows/artists via API
request - Overseerr request workflow
remote - Laptop→NAS management patterns

/servarr:servarr setup docker
/servarr:servarr integrate
/servarr:servarr add movie "Inception"
/servarr:servarr remote

Full documentation

immich

Deploy and manage Immich, a self-hosted photo and video management solution:

setup - Docker Compose deployment guide
configure - Environment variables and settings
backup - Database and filesystem backup/restore
library - External library setup for existing photo directories
upload - CLI bulk upload with album creation
api - REST API usage with curl examples
ml - Machine learning configuration (CLIP, faces, OCR)
mobile - Mobile app setup and backup

/immich:immich setup
/immich:immich backup
/immich:immich upload /path/to/photos
/immich:immich library /mnt/nas/photos

Full documentation

glinet

Configure and manage GL.iNet routers (firmware v4.x):

setup - First time setup and internet configuration
vpn - WireGuard, OpenVPN, Tailscale client/server
adguard - Built-in AdGuard Home ad blocking
gateway - Drop-in gateway for existing networks
dns - Encrypted DNS (DoH, DoT, DNSCrypt)
multiwan - Failover and load balancing
firewall - Port forwarding, DMZ, access control
storage - USB/Samba/WebDAV/DLNA

/glinet:glinet setup
/glinet:glinet vpn wireguard
/glinet:glinet adguard
/glinet:glinet gateway

Full documentation

umami

Deploy and manage Umami, a privacy-focused open-source web analytics platform:

setup — Docker Compose, source, and cloud installation
api — @umami/api-client TypeScript client for all endpoints
track — Client-side and server-side event tracking
stats — Website statistics, metrics, and active users
reports — Attribution, funnel, retention, journey, revenue, UTM
realtime — Live visitor data
teams — Team and user management

/umami:umami setup
/umami:umami api getWebsites
/umami:umami track purchase
/umami:umami stats <websiteId>
/umami:umami reports funnel

Full documentation

flue

Expert on Flue — the open-source TypeScript framework (by the Astro team) for building durable, autonomous AI agents and workflows on the Pi harness:

quickstart — Install, first agent, run locally
agent — Building agents with createAgent (harness-driven)
workflow — Authoring deterministic run(...) workflows
channels — Verified inbound provider events (Slack, GitHub, Stripe, …)
cli — init, dev, run, connect, build, add, update, logs, docs
sdk — createFlueClient, agents, workflows, runs, events
api — Agent, routing, provider, sandbox, and channel APIs
deploy — Cloudflare, AWS, Docker, Fly, Node, Railway, Render, SST, and more
sync/diff — Stay updated with upstream docs

/flue:flue quickstart
/flue:flue agent
/flue:flue deploy cloudflare
/flue:flue add channel slack

Full documentation

pi

Expert on Pi (pi.dev, @earendil-works/pi-coding-agent) — the minimal terminal coding harness. Procedure for extending Pi, improving it, adding custom providers, writing extensions / skills / themes / pi packages, forking, and running from source:

extend / loop — packages/agent (pi-agent-core)
provider — packages/ai (pi-ai), custom providers
extension / skill / theme / package — packages/coding-agent
fork — read docs/development.md Forking / Rebranding
source — npm install --ignore-scripts, npm run check, ./test.sh, ./pi-test.sh
sync/diff — Stay updated with upstream docs

/pi:pi extend
/pi:pi provider
/pi:pi extension
/pi:pi source

Full documentation

wealthfolio

Expert on Wealthfolio — the open-source, private, local-first portfolio & net-worth tracker (desktop, iOS, self-hosted Docker):

start — Install & onboarding (download, accounts, first activities)
concepts — Tracking modes, activity types, cost basis & lots, performance metrics
import — CSV import: column mapping, activity types, troubleshooting
guide — Dashboards, accounts, spending & budgets, goals, retirement/FIRE, contribution limits
connect — Wealthfolio Connect broker sync & encrypted multi-device sync
ai — Local AI assistant via Ollama
self-host — Docker, Docker Compose, Unraid, Proxmox, Coolify, reverse proxy
addon — Addon development (getting started, API reference)
sync/diff — Stay updated with upstream docs

/wealthfolio:wealthfolio start
/wealthfolio:wealthfolio import
/wealthfolio:wealthfolio self-host docker
/wealthfolio:wealthfolio addon getting-started

Full documentation

1password-cli

Expert on the 1Password CLI (the op command) for managing 1Password from the terminal and loading secrets into scripts and CI:

start — Install (macOS/Windows/Linux), enable desktop-app integration, first sign-in
signin — Manual sign-in, multiple accounts, biometric unlock
secrets — Secret references (op://vault/item/field), op read
run / inject — op run -- <cmd> and op inject for env vars, .env, and config files
item / vault — Full CRUD for items and vaults, fields, share links, permissions
service-account — Automation/CI auth via OP_SERVICE_ACCOUNT_TOKEN; Connect server
plugin — Shell plugins (op plugin) for 80+ third-party CLIs
ssh — 1Password SSH agent, SSH keys, git commit signing
sync/diff — Stay updated with upstream docs

/1password-cli:1password-cli start
/1password-cli:1password-cli secrets
/1password-cli:1password-cli run
/1password-cli:1password-cli service-account

Full documentation

portainerctl

Expert on portainerctl — Portainer's official CLI for driving Portainer Business Edition over its REST API:

install — Binary download / build from source; supported Portainer version
auth — API-token (PAT) auth, kubeconfig-style contexts, PORTAINERCTL_* env vars for CI
environment — Environments/endpoints, environment groups, tags, snapshots
stack — Deploy Compose/Swarm/Kubernetes stacks from file or Git (GitOps), lifecycle
docker / kubernetes — Containers/images/volumes/networks, Helm, kubectl & docker passthrough
edge — Edge stacks, groups, jobs, configs, update schedules
admin — Users/teams/RBAC, registries, webhooks, backups, licensing, observability
sync/diff — Stay updated with upstream README

/portainerctl:portainerctl install
/portainerctl:portainerctl auth
/portainerctl:portainerctl stack deploy
/portainerctl:portainerctl environment list

Full documentation

beads

Expert on beads (bd) — a dependency-aware, Dolt-backed issue tracker built for AI coding agents:

concepts — beads as a work graph, hash IDs that never collide, what bd ready computes
deps — blocking vs annotation-only dependency types, trees, cycles
workflows — formula → proto → molecule, plus gates and ephemeral wisps
sync — bd dolt push/pull over refs/dolt/data, bootstrap, and what JSONL export is not
setup — bd setup claude, the SessionStart bd prime hook, other IDE recipes
json — the stable --json output contract for agents
version — installed bd vs latest release, and the docs-vs-release drift
sync/diff — stay updated with upstream docs

/beads:beads quickstart
/beads:beads ready
/beads:beads workflows
/beads:beads version

Full documentation

tower

How to run a control tower over a fleet of herdr agents — the operating protocol for the agent that drives the others, including the acceptance gates that decide when a job is done:

status / catalog — Session start reads the work graph and the status board; catalog names entitled vs missing kinds
unpaid ask — settle every prior owner ask before starting a new job; a pane start does not pay a question
dispatch — the spec is a file, not a prompt; pane labels; start argv per kind; land-checking that the prompt actually arrived
planes — herdr is the default; inline is one cheap fact to fill the spec
watch/verify — idle is not done: a named marker file on disk plus your own re-check of the central claim against gh, the disk, or the DB
reconvene — after a marker: verify, then one owner table (Project, Status, What is true, Wrong/gap, You)
staff — PM · mentor · worker · adversary · reviewer · scout · QA, and which of those a given task actually requires
layout — one tab per checkout; an extra tab means an extra worktree and branch
worktree — default implement seat: list, then open or create; do not write on home main
escalate — re-verify, then an owner-queue item, or decline because it is already fixed
handoff — rewrite a session snapshot from live herdr and markers; a handoff note is not live state
close — retiring agents, one-shot panes, when a workspace may close, and reloading a plugin without tearing down a space
graph — work graph file, unpaid ask, node states, and typed edges
pitfalls — the failure catalogue, each entry with what it actually cost
gates — a gates file of CHECK / EXPECT / EVIDENCE outcomes verified by the vendored zero-dependency checker (Leonxlnx/unlazy, MIT)
scripts — tower-watch.sh / tower-poke.sh, the marker-watch and poke loop; gate-check.mjs

/tower:tower status
/tower:tower catalog
/tower:tower dispatch
/tower:tower reconvene
/tower:tower staff
/tower:tower escalate
/tower:tower handoff
/tower:tower worktree
/tower:tower gates status gates/job.md

Full documentation

pr-standard

The house standard for GitHub pull request descriptions. One card an agent or a human reads before writing a PR body:

rules — the six ASD-STE100 rules: one idea per sentence, the same word for the same thing, active voice, imperative in steps, no filler, exact technical names
template — Labels checklist, Summary, How It Works (mermaid), Linked issues, Testing evidence, Notes
labels — the four-pack: type, priority:*, t-shirt:*, and that repo's own area:*
diagram — which diagram a change needs, when to commit a .mmd, and the three mermaid patterns GitHub refuses to render
fork — why the conversation that wrote the code must not write the body

/pr-standard:pr-standard write
/pr-standard:pr-standard fix 812
/pr-standard:pr-standard labels
/pr-standard:pr-standard diagram

Hard rule for agents: fill the template. Never gh pr create -b "<one paragraph>".

Full documentation

skill-release-preflight

Three commands before every push to this repo. Each takes under a second, and each has already caught a failure that reached CI:

version — release-please owns sync.json and plugin.json and bumps them in lockstep; a hand edit desyncs them and fails CI. Revert to .release-please-manifest.json, never bump the other file to match
register — the six registration points, and EXEMPT_SYNC for a skill with no upstream
new-skill — the Release-As: 1.0.0 footer, without which a new skill ships as 1.1.0 and its changelog links to a tag that does not exist
check — every MISMATCH: line check-consistency.sh can print, and the one WARNING: that is expected
preflight — consistency, gh label list, git status --porcelain, then push

/skill-release-preflight:skill-release-preflight preflight
/skill-release-preflight:skill-release-preflight check
/skill-release-preflight:skill-release-preflight version

Full documentation

gh-issue-tracker

Install, configure, and manage gh-issue-tracker — lightweight error tracking that creates GitHub Issues:

setup — guided setup for any framework (Next.js, Express, generic Node.js)
architecture — server-only vs client+server decisions
security — recommendations and proxy patterns
verify — verification and troubleshooting
triage — issue triage and management workflows

Full documentation

obsidian

Manage and optimize Obsidian vaults — beyond syntax, into systems:

vault — folder structures, naming, MOCs, bulk operations, orphan detection
dataview — DQL queries, tables, lists, task queries, JavaScript API
templater — dynamic templates, JS execution, file creation, date handling
workflows — Zettelkasten, PARA, daily notes, periodic reviews
plugins — Dataview, Templater, Tasks, Calendar, Excalidraw, and more
ai / mcp — MCP servers, Local REST API, agent workflows
sync / publish — Obsidian Sync, git-based sync, Obsidian Publish

Full documentation

adguard

Deploy, configure, and manage AdGuard Home — network-wide DNS ad blocking and filtering:

setup — Docker and bare-metal deployment
configure — DNS filtering, blocklists, upstreams
encryption — DoH, DoT, DoQ, DNSCrypt
dhcp — DHCP server setup
clients — per-client rules and settings
api — REST API automation

/adguard:adguard setup docker
/adguard:adguard configure

Full documentation

apprise-api

Deploy, configure, and drive the Apprise API — a self-hosted notification gateway that turns one HTTP request into notifications across 100+ services:

deploy — Docker, Docker Compose, hardened rootless Kubernetes
configure — every APPRISE_* environment variable, TEXT and YAML config
notify — stateless (/notify) and stateful (/notify/{KEY}) requests
key — save, read, and delete configuration keys
tags — AND/OR tag expressions to route to a subset of saved URLs
attach — file upload, remote URL, and JSON attachment forms
webhook — remap third-party JSON payloads onto title/body
troubleshoot — response codes, error lookup, URL escaping

/apprise-api:apprise-api deploy
/apprise-api:apprise-api configure
/apprise-api:apprise-api key my-alerts
/apprise-api:apprise-api troubleshoot 424

Full documentation

Adding New Skills

Create a new directory under skills/:

skills/
└── my-new-skill/
    ├── .claude-plugin/
    │   └── plugin.json
    ├── commands/
    ├── skills/
    ├── sync.json           # Required for CI automation
    └── README.md

Create sync.json for automated updates:

{
  "name": "my-new-skill",
  "version": "1.0.0",
  "description": "What it does",
  "sources": [
    {
      "url": "https://example.com/docs.md",
      "target": "skills/my-skill/docs/readme-upstream.md",
      "freshness_days": 14
    }
  ],
  "cache_dir": ".cache"
}

Add entry to .claude-plugin/marketplace.json:

{
  "name": "my-new-skill",
  "source": "./skills/my-new-skill",
  "description": "What it does",
  "version": "1.0.0"
}

Add skill to CI sync array in .github/workflows/sync-docs.yml if it has upstream docs
Commit and push

CI Automation

The repository includes automated workflows for keeping skill documentation up to date.

Scheduled Sync

Runs bi-weekly (1st and 15th of each month) to check for upstream documentation changes. Only syncs skills with actual upstream documentation sources.

Manual trigger: Actions > Sync Skill Documentation > Run workflow

Options:

force: Force sync even without detected changes
dry_run: Check for changes without creating PR

Skills with CI sync enabled:

umbrel-app, claude-code-expert, agent-browser, chat-sdk, ai-sdk, agent-skills, hermes, honcho, firecrawl, servarr, obsidian, adguard, immich, glinet, umami, flue, wealthfolio, 1password-cli, portainerctl, x-engagement, beads, tower, pi

Automated Releases (release-please)

This repository uses release-please to automate versioning and releases based on Conventional Commits.

How it works:

Push commits using conventional format: feat(skill-name): description
Release-please detects changes and creates a release PR
The PR updates versions and CHANGELOG automatically
When merged, creates a GitHub release with tag (e.g., umbrel-app-v1.3.0)

Commit types:

Prefix	Version Bump	Changelog Section
`feat`	Minor (1.x.0)	Features
`fix`	Patch (1.0.x)	Bug Fixes
`docs`	Patch (1.0.x)	Documentation
`chore`	None	Hidden

Configuration files:

release-please-config.json - Release settings and package definitions
.release-please-manifest.json - Current versions for each skill

Configuration

Set custom schedule via repository variable:

SYNC_SCHEDULE: "0 6 1,15 * *"  # Cron format

Structure

skills/
├── .claude-plugin/
│   └── marketplace.json          # Marketplace manifest
├── .github/
│   └── workflows/
│       ├── scripts/
│       │   ├── check-consistency.sh # Registration consistency checks
│       │   ├── sync-skill.sh        # Generic sync script
│       │   └── sync-marketplace.sh  # Marketplace sync script
│       ├── release-please.yml       # Automated releases
│       ├── sync-docs.yml            # Scheduled sync workflow
│       ├── sync-hermes-on-release.yml
│       ├── sync-pi-on-release.yml
│       ├── sync-marketplace.yml
│       └── validate.yml             # CI validation
├── skills/
│   ├── umbrel-app/               # Umbrel app development
│   ├── claude-code-expert/       # Claude Code knowledge base
│   ├── agent-browser/            # Browser automation for AI agents
│   ├── chat-sdk/                 # Vercel Chat SDK
│   ├── ai-sdk/                   # Vercel AI SDK
│   ├── agent-skills/             # Agent Skills specification
│   ├── hermes/                   # Hermes Agent self-knowledge
│   ├── honcho/                   # Honcho AI-native memory platform
│   ├── safe-delete/              # Safe file deletion
│   ├── x-engagement/             # X/Twitter engagement
│   ├── gh-issue-tracker/         # Error tracking via GitHub Issues
│   ├── firecrawl/                # Firecrawl web scraping API
│   ├── servarr/                  # Media stack (*arr suite + Plex)
│   ├── obsidian/                 # Obsidian vault management
│   ├── adguard/                  # AdGuard Home DNS filtering
│   ├── immich/                   # Immich photo/video management
│   ├── glinet/                   # GL.iNet router management
│   ├── umami/                    # Umami web analytics
│   ├── flue/                     # Flue open agent framework
│   ├── pi/                       # Pi coding harness (extend / improve)
│   ├── wealthfolio/              # Private local-first portfolio tracker
│   ├── 1password-cli/            # 1Password CLI (op) secrets management
│   ├── portainerctl/             # Portainer CLI (Business Edition)
│   ├── beads/                    # Beads (bd) dependency-aware issue tracker
│   ├── tower/                    # Control tower + acceptance gates over a herdr fleet
│   ├── pr-standard/              # House standard for GitHub PR descriptions
│   └── skill-release-preflight/  # Pre-push checks for this repo
└── README.md

Agent Skills Compatibility

All skills in this repository follow the Agent Skills open format specification:

Requirement	Status
Valid `name` field (lowercase, hyphens)	✓
Name matches directory	✓
Descriptive `description` with triggers	✓
SKILL.md under 500 lines	✓
Progressive disclosure pattern	✓
References in docs/ folder	✓

Specification Compliance

SKILL.md format: YAML frontmatter + Markdown body
Progressive disclosure: Metadata → Instructions → References (docs/)
Directory structure: skills/<name>/skills/<name>/SKILL.md
Token efficiency: Main skills ~100 lines, detailed docs loaded on-demand

Validate with skills-ref

pip install skills-ref
skills-ref validate ./skills/umbrel-app/skills/umbrel-app

For more information, see agentskills.io or the agent-skills skill.

License

MIT

Name		Name	Last commit message	Last commit date
.claude-plugin		.claude-plugin
.github/workflows		.github/workflows
skills		skills
.gitignore		.gitignore
.release-please-manifest.json		.release-please-manifest.json
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
README.md		README.md
release-please-config.json		release-please-config.json
skills.toml		skills.toml

Latest commit

History

Folders and files

Repository files navigation

AI Agent Skills

Available Skills

Installation

Recommended: install with zskills (Rust CLI package manager)

Inside Claude Code (built-in)

Or add to project settings manually

Usage

Slash Commands

Natural Language

Skill Details

umbrel-app

claude-code-expert

agent-browser

chat-sdk

ai-sdk

agent-skills

hermes

honcho

safe-delete

x-engagement

firecrawl

servarr

immich

glinet

umami

flue

pi

wealthfolio

1password-cli

portainerctl

beads

tower

pr-standard

skill-release-preflight

gh-issue-tracker

obsidian

adguard

apprise-api

Adding New Skills

CI Automation

Scheduled Sync

Automated Releases (release-please)

Configuration

Structure

Agent Skills Compatibility

Specification Compliance

Validate with skills-ref

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages