skills

Monorepo of Claude Code plugins and slash commands by Ricardo Arturo Cabral Mejía.

Plugins

Plugin	Description	Tools
actual-budget	Interact with your self-hosted Actual Budget instance — check balances, add transactions, and query budgets	Claude + Cursor
agent-resources	Build and audit Claude Code skills, hooks, MCP servers, and subagents	Claude + Cursor
audiobookshelf	Interact with a self-hosted Audiobookshelf instance — list libraries, browse and search library items, and check listening progress	Claude + Cursor
autoresearch	Autonomously optimize Claude Code skills using Karpathy's autoresearch methodology — binary evals, prompt mutation, and iterative improvement loops	Claude
brain	Workspace-wide knowledge graph backed by LatticeDB — syncs a graph-json snapshot in, defaulting to graphify's output (`learn_from`), answers questions over it via Cypher (`recall`), writes/links a single fact directly (`remember`), soft- or permanently deletes a node/edge (`forget`), and reports staleness/cost-to-re-study for any previously-learned path (`study_status`)	Claude + Cursor
consider	Apply a decision-making framework — inversion, first-principles, second-order effects, Pareto, via negativa, opportunity cost, or Eisenhower prioritization — to a real decision or tradeoff moment	Claude + Cursor
cronjobs	Schedule recurring or one-time jobs using natural language — 'every 3 minutes', 'every weekday at 9am', 'once in 5 minutes'	Claude
container-management	Maintain Docker Compose services and custom images — update base images, pin sha256 digests, manage Containerfile/Dockerfile dependencies, test builds, and log all changes	Claude + Cursor
doubt-driven-development	Adversarial fresh-context review of non-trivial in-flight decisions before they stand — CLAIM, EXTRACT, DOUBT, RECONCILE, STOP — with optional user-authorized cross-model escalation	Claude
elevenlabs	Generate speech, transcribe audio, create music and sound effects, and build voice agents using the ElevenLabs API	Claude + Cursor
executable-skepticism	Verification protocol that turns a theory, paper, model, or confident quantitative claim into a falsifiable, runnable test instead of a debate in prose	Claude
finance-manager	Reconcile bank statements against ActualBudget, run household financial reviews (net worth, goal tracking, optimization), look up Bitcoin transactions/addresses/wallet descriptors via mempool.space, and onboard or manage the plugin's tracked accounts, wallets, and periodic sync jobs	Claude + Cursor
github-manager	Autonomous GitHub repository manager — handles webhook events for issues, PRs, discussions, CI failures, and security alerts	Claude
home-assistant	Interact with Home Assistant via the REST API — get entity states, call services, fire events, render Jinja2 templates, and query state history	Claude + Cursor
jj	Use Jujutsu (jj) instead of git for version control — core concepts plus actionable skills for everyday operations (commit, split, squash, cleanup, checkout, rebase)	Claude
journal	Keeps a series of narrative journals about what you've been doing, written from Claude's own perspective, by reading session history and memory	Claude
knowledge-wiki	Maintains a self-updating, cross-linked markdown knowledge base — ingests durable facts and research as topic pages instead of re-deriving them each time, with query and link-consistency checks	Claude
lightning	Bitcoin Lightning payment tools (parse/pay invoices, balances, transactions) via the official Alby MCP server over Nostr Wallet Connect	Claude
nats	Connect Claude Code agents over NATS — message, ping, and discover other agents point-to-point	Claude + Cursor
netshoot	Network troubleshooting inside Docker container networks using nicolaka/netshoot	Claude
nostr	Nostr channel for Claude Code — decentralized messaging over Nostr relays with DM pairing, allowlists, relay pool management, and NIP-04 encrypted DMs	Claude
paperless	Upload documents to and search a Paperless-ngx instance via its REST API	Claude + Cursor
replicator	Grows and prunes this instance's own skill set — captures reusable procedures during work, and nightly meditates on usage history and frontier sources to build or mute skills with scrutiny; publishes a gene registry over Nostr and mirrors it to a GitHub gist	Claude
repo-hygiene-sweep	Sweep every standalone repo in a multi-repo workspace for uncommitted or unpushed work, without missing the ones a plain `git status`/`jj status` from the workspace root can't see	Claude
research-tools	Deliberately-invoked research skills — competitive analysis, deep-dive investigation, feasibility checks, landscape mapping, options comparison, and technical implementation research	Claude
sandbox-manager	Manage the Claude Code sandbox itself — restart sessions, manage its own plugins/marketplaces, and check post-restart health	Claude
simple-english	Write or rewrite technical text with the rules of ASD-STE100 Simplified Technical English so it is clear, unambiguous, and free of AI slop	Claude
technitium-dns	Manage a self-hosted Technitium DNS Server — zones, records, stats, and cache	Claude + Cursor
telegram	Telegram channel for Claude Code — messaging bridge with built-in access control, pairing, and full Bot API coverage including voice note transcription	Claude
telegram-ng	Telegram channel for Claude Code — messaging bridge with built-in access control. Fork of Anthropic's official telegram plugin for local development	Claude
wallabag	Save, search, and manage read-it-later articles via your Wallabag instance	Claude + Cursor
webhooks	Receive webhook events from external systems as channel notifications — HMAC-SHA256, IP allowlisting, BullMQ processing	Claude

actual-budget

Skill	Description
`/actual-budget:access`	Set up Actual Budget credentials — save the server URL and password
`/actual-budget:query-budget`	Query accounts, check balances, view recent transactions, and trigger bank sync

agent-resources

Skill	Description
`/agent-resources:create-agent-skills`	Expert guidance for creating, writing, building, and refining Claude Code Skills
`/agent-resources:create-hooks`	Create Claude Code hooks (PreToolUse, PostToolUse, Stop, SessionStart, UserPromptSubmit)
`/agent-resources:create-mcp-servers`	Expert guidance for building MCP servers for Claude integrations (Python/TypeScript)
`/agent-resources:create-subagents`	Expert guidance for creating, building, and using Claude Code subagents
`/agent-resources:audit-skill`	Audit a SKILL.md file for YAML compliance, pure XML structure, progressive disclosure, and best practices
`/agent-resources:audit-subagent`	Audit a subagent configuration file for role definition, prompt quality, and tool selection
`/agent-resources:heal-skill`	Apply corrections to a skill's SKILL.md based on mistakes discovered during execution, with approval workflow

audiobookshelf

Skill	Description
`/audiobookshelf:access`	Set up Audiobookshelf credentials — save the server URL and API key
`/audiobookshelf:query-library`	List libraries, browse or search items, view item details, check listening progress
`/actual-budget:add-transaction`	Add a transaction — spending, income, or any financial event

brain

Skill	Description
`/brain:learn-from`	Sync a graph-json snapshot into the brain (the workspace's LatticeDB knowledge graph) — defaults to graphify's `graphify-out/graph.json`, or pass a `path` to sync any file in the same schema — creates new nodes/edges, updates changed ones, deletes ones no longer present
`/brain:recall`	Query the brain — ask a plain-language question about the workspace's code/docs/concepts, or run a raw Cypher query directly
`/brain:remember`	Write a single fact into the brain, optionally linked to existing nodes by gid or search string
`/brain:forget`	Soft (default, recoverable) or permanent delete of a node or edge from the brain
`/brain:study-status`	Report whether a previously-learned path is stale and roughly what re-studying would cost, without triggering a re-study — shells out to graphify's own `detect_incremental()`

finance-manager

Skill	Description
`/finance-manager:setup-finance-manager`	Onboard the plugin for first use (household, accounts, hot/cold wallets, ownership, connecting Actual Budget/Paperless-ngx) or review/add/remove tracked entries and periodic sync jobs
`/finance-manager:reconcile-statement`	Reconcile a bank statement against ActualBudget — syncs accounts, matches transactions, self-improves reconciliation rules
`/finance-manager:manage-paperless-workflows`	Create or fix Paperless-ngx workflows so bank statement documents auto-tag correctly
`/finance-manager:query-mempool`	Look up Bitcoin transactions, addresses, and wallet descriptor (single-sig or multisig) balances/history via the mempool.space API
`/finance-manager:review-finances`	Run a household financial review — execution audit, net worth/liquidity check, tax/expense optimization scan, and a single-sitting report with goal tracking and up to 3 next actions

github-manager

Skill	Description
`github-manager:manage-issues`	Handles GitHub issue events; triages labels, prompts for details, escalates external issues via Telegram
`github-manager:manage-pull-requests`	Handles GitHub PR events; auto-merges Dependabot patches, escalates external PRs via Telegram
`github-manager:manage-discussions`	Handles GitHub discussion events; silently monitors trusted users, escalates external discussions via Telegram
`github-manager:manage-ci`	Handles GitHub CI events; alerts on failures via Telegram
`github-manager:manage-projects`	Handles GitHub Projects v2 events; notifies on lifecycle changes, escalates external activity via Telegram
`github-manager:manage-admin`	Handles GitHub security alerts, collaborator changes, pushes, and admin events
`github-manager:create-stories`	Parses a PRD and creates a structured hierarchy of GitHub issues (stories + sub-issues) with milestone assignment and project board integration

journal

Skill	Description
`journal:update-journal`	Reads session activity since the last update across every project on this machine plus the memory system, judges whether it continues the current journal's cycle or starts a new one, and writes/closes entries accordingly

knowledge-wiki

Skill	Description
`knowledge-wiki:maintain-wiki`	Ingests durable facts/research into cross-linked markdown topic pages, answers lookups against the wiki, and checks it for broken links and orphaned pages

elevenlabs

Skill	Description
`elevenlabs:text-to-speech`	Convert text to speech in 70+ languages using ElevenLabs voice AI
`elevenlabs:speech-to-text`	Transcribe audio/video to text using ElevenLabs Scribe v2
`elevenlabs:agents`	Build real-time voice AI agents and assistants
`elevenlabs:music`	Generate instrumental tracks, songs, and background music from prompts
`elevenlabs:sound-effects`	Generate sound effects, ambient sounds, and audio textures from text
`elevenlabs:access`	Configure an ElevenLabs API key (ELEVENLABS_API_KEY)
`elevenlabs:elevenlabs-transcribe`	Batch or realtime audio transcription via CLI scripts

executable-skepticism

Skill	Description
`executable-skepticism:executable-skepticism`	Operationalize a claim, register numeric predictions before running any code, execute deterministically, then score every prediction pass/fail — failures first, derived-vs-installed called out

doubt-driven-development

Skill	Description
`doubt-driven-development:doubt-driven-development`	Adversarial fresh-context review of a non-trivial in-flight decision — CLAIM, EXTRACT, DOUBT, RECONCILE, STOP — with optional user-authorized cross-model escalation

nats

Skill	Description
`/nats:access`	Configure the NATS server URL and this agent's display name
`/nats:show-nats-status`	Show connection info, display name, and all discovered agents
`/nats:discover-agents`	Broadcast "who's there?" and list all discovered agents
`/nats:ping-agent`	Liveness check against one known agent, reports round-trip time
`/nats:send-message`	Send a free-form message directly to another agent

paperless

Skill	Description
`/paperless:access`	Save the instance URL, username, and password; verify connection
`/paperless:search-documents`	Full-text search, similarity search, or autocomplete
`/paperless:upload-document`	Upload a local file with optional metadata
`/paperless:view-content`	Display the full OCR-extracted text of a document by ID
`/paperless:view-document`	Download the archived PDF; when called from Telegram, sends the file to chat

sandbox-manager

Skill	Description
`sandbox-manager:restart-session`	Fires automatically on a `/clear` channel message; sends `/clear` + Enter to the tmux pane running this Claude Code session
`sandbox-manager:rename-session`	Names or renames the current session; fires on a `/rename <name>` channel message
`sandbox-manager:resume-session`	Restores a named session, replacing the current conversation; fires on a `/resume <name>` channel message
`sandbox-manager:branch-session`	Branches the current conversation at the current point; fires on a `/branch` channel message
`sandbox-manager:compact-session`	Compacts the current conversation (optionally with retention instructions); fires on a `/compact` channel message
`sandbox-manager:background-session`	Hands the current work to a background agent, freeing the interactive pane; fires on a `/background` channel message
`sandbox-manager:export-session`	Exports the current conversation to a file; fires on a `/export [path]` channel message
`sandbox-manager:exit-session`	Hard-exits the session (vs. `/clear`'s soft reset); fires on a `/exit` channel message
`sandbox-manager:manage-plugins`	Adds plugin marketplaces, or installs/updates/enables/disables/uninstalls a plugin for this sandbox's own running session
`sandbox-manager:setup-hooks`	Installs the curated hook set (rm guard, channel-reply enforcement, usage alerts, idle tracker, session handoff reader, new-session notifier, statusline cache wrapper, and more)
`sandbox-manager:check-login-expiry`	Checks whether this session's login is about to expire and reports over Telegram if so; fires on a daily cronjobs job
`sandbox-manager:post-restart-check`	Post-restart health check — SSH signing key still has its public half, Docker Buildx present, no Containerfile pip package silently dropped by a rebuild
`sandbox-manager:whats-next`	Writes a comprehensive `whats-next.md` handoff document so work can resume with zero information loss after a context reset or restart
`sandbox-manager:add-to-todos`	Adds an item to `TO-DOS.md` with full context from the conversation, checking for near-duplicates first
`sandbox-manager:check-todos`	Lists outstanding items from `TO-DOS.md` and helps pick one to work on next

cronjobs

Skill	Description
`/cronjobs:cronjob`	Schedule a recurring or one-time job using natural language; fires channel notifications when due

simple-english

Skill	Description
`simple-english:simple-english`	Write or rewrite technical text per ASD-STE100 — classifies procedural vs. descriptive, applies the 53-rule catalog, runs a mandatory self-check before delivering

technitium-dns

Skill	Description
`/technitium-dns:access`	Save the server URL and API token (or username/password)
`/technitium-dns:query-dns-stats`	Query DNS stats — top clients, top domains, query counts, cache info
`/technitium-dns:manage-dns-zones`	List, create, delete, enable, or disable DNS zones
`/technitium-dns:manage-dns-records`	Add, list, update, or delete A, AAAA, CNAME, MX, TXT, SRV records
`/technitium-dns:manage-blocking`	Block or allow domains — check status, add/remove domain overrides, manage block list URLs

wallabag

Skill	Description
`/wallabag:access`	Save the instance URL and OAuth credentials
`/wallabag:save-url`	Save a URL to Wallabag to read later

webhooks

Skill	Description
`webhooks:receive-webhooks`	Configure webhook endpoints (add/edit/remove/list), set auth mode, manage IP allowlists, and react to inbound events

autoresearch

Skill	Description
`/autoresearch:optimize-skill`	Improve a SKILL.md using binary evals and iterative prompt mutation — use when a skill has reliability issues or produces inconsistent results

container-management

Skill	Description
`/container-management:container-management`	Update Docker base images with sha256 pinning, manage Containerfile dependencies, test builds, and keep an audit log

home-assistant

Skill	Description
`/home-assistant:access`	Configure Home Assistant credentials — save the server URL and long-lived access token
`/home-assistant:get-state`	Get the current state of one or all entities
`/home-assistant:set-state`	Create or update the state of an entity
`/home-assistant:call-service`	Call a service to control devices or trigger automations
`/home-assistant:fire-event`	Fire a custom event to trigger event-driven automations
`/home-assistant:query-history`	Query state history or logbook for one or more entities
`/home-assistant:render-template`	Render and debug a Jinja2 template

jj

Skill	Description
`jj:working-with-jj`	Use Jujutsu (jj) instead of git for all version control operations in this workspace — commits, history, branching, pushing
`/jj:commit`	Describes the current working-copy change (jj's equivalent of `git commit`), optionally stacking a new empty change on top
`/jj:split`	Splits the current (or a specified) change into two, by file/path or interactively — jj's equivalent of `git add -p && git commit`
`/jj:squash`	Moves changes from one change into another (usually working copy into parent) — jj's equivalent of `git commit --amend` or a fixup squash
`/jj:rebase`	Moves changes onto a different parent — jj's equivalent of `git rebase`, without stopping for conflicts
`/jj:checkout`	Switches to an existing change or bookmark — jj's equivalent of `git checkout`/`git switch`
`/jj:cleanup`	Removes stray empty changes and stale/orphaned workspaces

nostr

Skill	Description
`nostr:nostr`	Handle inbound Nostr DMs, send replies, publish notes, fetch events, and check relay status
`/nostr:configure`	Set up the Nostr channel — save the nsec, manage relays, configure subscribed event kinds
`/nostr:access`	Manage Nostr channel access — pairings, allowlists, policy
`/nostr:profile`	View or update the Nostr profile (kind:0 metadata)
`/nostr:relay-list`	Manage the NIP-65 relay list for this identity
`/nostr:fetch-event`	Fetch a Nostr event by ID or filter
`/nostr:verify-event`	Verify an event's schema, ID hash, and Schnorr signature
`/nostr:react`	React to an event with a NIP-25 reaction
`/nostr:bech32`	Encode/decode NIP-19 bech32 entities (note1, npub1, nevent1, etc.)
`/nostr:mine-pubkey`	Mine a vanity or proof-of-work Nostr keypair with rana

Skill	Description
`telegram:telegram`	Handle inbound Telegram messages, send replies, react, edit messages, download attachments, process voice notes
`/telegram:configure`	Save the bot token and review access policy
`/telegram:access`	Manage Telegram channel access — pairings, allowlists, DM/group policy

telegram-ng

Fork of Anthropic's official telegram plugin — see telegram-ng/README.md. This is the live channel driver for this workspace; telegram above is the unforked predecessor, kept in the repo for reference/other users.

Skill	Description
`/telegram-ng:configure`	Save the bot token and review access policy
`/telegram-ng:access`	Manage Telegram channel access — pairings, allowlists, DM/group policy
`telegram-ng:bot-api-reference`	Reference for Telegram's Bot API surface — consulted before modifying server.ts to add or change Telegram behavior

lightning

MCP-only — no SKILL.md skills. Ships an MCP server (@getalby/mcp) wired to a self-hosted NWC wallet; use its parse_invoice/pay_invoice/balance/transaction tools directly.

replicator

Skill	Description
`replicator:capture`	Queue a candidate skill the moment you notice something reusable was just learned during a session
`replicator:meditate`	Run the nightly replicator cycle — review gene usage, look inward/outward for skills worth building, build with scrutiny, prune stale ones, write a trace

repo-hygiene-sweep

Skill	Description
`repo-hygiene-sweep:repo-hygiene-sweep`	Checks every repo in a multi-repo workspace — the workspace repo plus every standalone repo listed in its root CLAUDE.md — for uncommitted or unpushed work

consider

Skill	Description
`/consider:consider`	Apply a decision-making framework to a real choice, tradeoff, or risky plan

research-tools

Skill	Description
`/research-tools:competitive`	Research the competitive landscape for a product or feature — who else solves this, how, and where the gaps are
`/research-tools:deep-dive`	Comprehensive, multi-source investigation of a topic — how it works, why it exists, limitations, current trends
`/research-tools:feasibility`	Honest reality check — can this actually be done given technical, resource, and external constraints
`/research-tools:landscape`	Map a domain's players, tools, trends, and gaps
`/research-tools:options`	Structured side-by-side comparison of options with a recommendation
`/research-tools:technical`	Research implementation approaches, libraries, and patterns with honest tradeoffs

Commands

Command	Description
`/create-skill`	Scaffold a new Claude Code skill plugin from scratch
`/dontforget`	Consolidate `remember/` notes into the persistent memory system
`/forget`	Triage and prune stale entries from `remember/`
`/unsubscribe`	Unsubscribe from a newsletter or mailing list using an unsubscribe URL
`/unfurl`	Resolve a minified or tracking URL to its final destination

Setup

1. Clone the repo

git clone git@github.com:cameri/skills.git ~/Workspace/skills

2. Install slash commands

Copy (or symlink) the commands into your Claude config directory so they are available in every session:

cp ~/Workspace/skills/commands/*.md ~/.claude/commands/

Or as symlinks so changes in the repo are picked up automatically:

for f in ~/Workspace/skills/commands/*.md; do
  ln -sf "$f" ~/.claude/commands/"$(basename "$f")"
done

3. Register the plugin marketplace

Run this once inside any Claude Code session:

/plugin marketplace add ~/Workspace/skills

4. Install plugins

/plugin install actual-budget@cameri-skills
/plugin install elevenlabs@cameri-skills
/plugin install github-manager@cameri-skills
/plugin install nats@cameri-skills
/plugin install paperless@cameri-skills
/plugin install cronjobs@cameri-skills
/plugin install technitium-dns@cameri-skills
/plugin install wallabag@cameri-skills

5. Reload plugins

/reload-plugins

After reloading, all plugin skills are available (e.g. /paperless:configure, /actual-budget:budget).

Installing for Cursor

Cursor reads the same SKILL.md format as Claude Code (frontmatter name/description, discovered from .cursor/skills/<name>/SKILL.md) and the same .mcp.json shape ({"mcpServers": {...}}, read from ~/.cursor/mcp.json). No plugin content needs to change for Cursor — only how it's discovered.

The 12 plugins marked "Claude + Cursor" in the table above (paperless, actual-budget, technitium-dns, home-assistant, wallabag, elevenlabs, nats, container-management, finance-manager, audiobookshelf, agent-resources, consider) work under Cursor. Channel plugins (telegram, telegram-ng, nostr, webhooks, cronjobs, sandbox-manager) stay Claude-only — they react to inbound background messages, which has no Cursor equivalent since Cursor is an interactive editor, not a background message host. netshoot also stays Claude-only for now.

1. Symlink each skill directory

Each plugin's skills live at <plugin>/skills/<skill-name>/SKILL.md. Symlink each one into ~/.cursor/skills/<plugin>-<skill-name> so Cursor can discover it:

for skill_dir in ~/Workspace/skills/paperless/skills/*/; do
  skill_name=$(basename "$skill_dir")
  ln -s "$skill_dir" ~/.cursor/skills/paperless-"$skill_name"
done

Repeat for each of the 9 Cursor-eligible plugins, substituting the plugin name in both the source path and the ~/.cursor/skills/ prefix.

2. Register MCP servers (only plugins with one)

Most of the 9 Cursor-eligible plugins are skill-only (the SKILL.md drives curl/API calls directly) and need no MCP server entry — Step 1 alone is enough. nats is the exception: it ships a real MCP server (nats/.mcp.json). Paste its contents into ~/.cursor/mcp.json, replacing ${CLAUDE_PLUGIN_ROOT} (a Claude Code–only variable Cursor doesn't expand) with the plugin's absolute path:

{
  "mcpServers": {
    "nats": {
      "command": "bun",
      "args": ["run", "--cwd", "/absolute/path/to/skills/nats", "--shell=bun", "--silent", "start"]
    }
  }
}

If a plugin gains an .mcp.json in the future, apply the same pattern (copy the plugin's own .mcp.json into ~/.cursor/mcp.json's mcpServers object, resolving any ${CLAUDE_PLUGIN_ROOT} reference to an absolute path).

Alternative: install via `npx skills`

vercel-labs/skills is a separate community CLI that installs Claude Code skills straight from a GitHub repo, without registering a plugin marketplace. It works against this repo too.

Install every skill in this repo:

npx skills add cameri/skills

Install a specific plugin's skill directly (skills live at <plugin>/skills/<skill-name>):

npx skills add https://github.com/cameri/skills/tree/main/actual-budget/skills/access

Or pick skills by name out of the whole repo:

npx skills add cameri/skills -s access query-budget add-transaction

Useful flags: --list to see what's available before installing, -g/--global to install to your user directory instead of the current project, --copy to copy files instead of symlinking, and -a claude-code to target Claude Code if you have other supported agents installed. You can also run a skill without installing it:

npx skills use cameri/skills --skill access --agent claude-code

See the vercel-labs/skills README for the full command reference (list, find, update, remove, init).

License

MIT

FilesExpand file tree

README.md

Latest commit

History