An MCP server that jacks Agent Skills directly into your LLM's brain.
Tool search / deferred tools. By default, skilljack delivers its skill catalog via MCP server instructions, which arrive in the
initializehandshake and are visible to the model even when tool search / deferred tool loading is enabled (the default on modern Claude Code) — so automatic skill activation works out of the box. The legacy--catalog=tool-descriptionmode (not recommended — see Catalog channel below) delivers the catalog through theload-skilltool description instead, but tool-search clients defer tool descriptions out of context, so that mode needsENABLE_TOOL_SEARCH=falseto auto-activate.
npm install -g @skilljack/mcpOr run directly with npx:
npx @skilljack/mcp /path/to/skillsgit clone https://github.com/olaservo/skilljack-mcp.git
cd skilljack-mcp
npm install
npm run build
node dist/index.js /path/to/skills# Single directory
skilljack-mcp /path/to/skills
# Multiple directories
skilljack-mcp /path/to/skills /path/to/more/skills
# GitHub repository (allowlisted via GITHUB_ALLOWED_ORGS / _USERS)
GITHUB_ALLOWED_ORGS=acme skilljack-mcp github.com/acme/skills
# Well-known publisher (allowlisted via WELL_KNOWN_ALLOWED_ORIGINS).
# Each entry's SHA-256 digest is verified against the published index.
# Every fetched URL — the index, each artifact, and any redirect target — must
# match an allowlisted origin, so if a publisher serves artifacts from a
# separate CDN origin, add that origin to the list too (comma-separated).
WELL_KNOWN_ALLOWED_ORIGINS=https://example.com \
skilljack-mcp https://example.com/.well-known/agent-skills/
# Using environment variable
SKILLS_DIR=/path/to/skills skilljack-mcp
# Static mode (no file watching)
skilljack-mcp --static /path/to/skills
# Serve over HTTP instead of stdio (stateless Streamable HTTP on POST /mcp)
skilljack-mcp --http=3000 /path/to/skills
# or: SKILLJACK_HTTP_PORT=3000 skilljack-mcp /path/to/skillsTransports: stdio (default) or stateless HTTP (--http[=port], default 3000, or SKILLJACK_HTTP_PORT). HTTP serves the core skill surface (load-skill, skill-resource, skill:// resources, skills/list / skills/get, and the /skill and per-skill prompts) at POST /mcp. Use the --http=<port> form: --http 3000 treats 3000 as a skill directory. The HTTP server listens on all interfaces and has no authentication. Both transports use the MCP SDK's era-aware entry points, so a client negotiating MCP 2026-07-28 and a 2025-era client are served from the same server definition. Discovery-on-change works the same as stdio — file watchers and remote-source polling keep the skill state fresh, and every request (including each new client's initialize) reads it. Because the transport is stateless it does not push listChanged/resources/updated notifications: already-connected clients see changes on their next request or reconnect. The MCP-Apps configuration UI is stdio-only.
Skills extension (SEP-2640): the server declares io.modelcontextprotocol/skills (with directoryRead) through the MCP TypeScript SDK's skills extension and implements skills/list, skills/get and resources/directory/read. Each entry carries the skill's verbatim frontmatter and a {uri, digest, size} manifest of SKILL.md and every supporting file, so a host can verify what it reads. Files are served one at a time via resources/read under skill://<skill-path>/<file>; on 2026-07-28 connections the skills/list and skills/get results also carry the ttlMs / cacheScope caching attributes the extension requires. Supporting files larger than MAX_FILE_SIZE_MB (default 1) are not served and therefore not listed. skill://index.json is still served for clients written against the pre-v1 index shape.
Tools: --tools=<auto|always|never> (or SKILLJACK_TOOLS) controls the load-skill and skill-resource tools. auto (default) offers them unless the client declares io.modelcontextprotocol/skills in its capabilities, as MCP Inspector and MCPJam do; such a host loads skills itself through skills/list, skills/get and resources/read, so the tools are disabled after initialize. The catalog in server instructions lists each skill's skill:// URI for that path. Over stateless HTTP the initialize handshake and later requests hit different server instances, and a 2026-07-28 connection has no initialize handshake at all, so auto behaves as always in both cases; use never there if the host has the extension. With --catalog=tool-description the catalog lives in the tool description, so auto is ignored.
Catalog channel: --catalog=<instructions|tool-description> (or SKILLJACK_CATALOG) picks which single channel delivers the <available_skills> catalog — never both. instructions (default): server instructions, sent in the initialize handshake so it survives tool-search deferral; on stdio it is frozen at startup — skill changes need a restart (over HTTP, newly connecting clients always get a current catalog). tool-description is not recommended and is retained only as a legacy escape hatch (and as the control condition in the evals): the load-skill tool description is invisible to clients that defer MCP tool descriptions (e.g. Claude Code tool search, the modern default), so auto-activation silently breaks unless the client sets ENABLE_TOOL_SEARCH=false; and its one advantage — live catalog refresh via tools/listChanged — is a prompt-cache trap, because tool definitions sit at the top of the cached prompt prefix, so every refresh invalidates the entire cache (tools + system prompt + conversation history) and re-writes it at cache-write prices. Measured 2026-07-06 (#78): it has no cost advantage over instructions in steady state either.
This server comes along with a MCP Apps-based UI for clients that support it. Instead of fiddling with config files or environment variables, you can just configure your skills locations and skill visiblity directly in your chat window. Locations set in the UI are saved to ~/.skilljack/config.json and take effect only when no directories are given on the command line or in SKILLS_DIR. The UI is not available with --static or --http.
(Screenshots below are from Claude Desktop in dark mode.)
For complete documentation, just ask your assistant:
"how do I use skilljack?" or "how does skilljack work behind the scenes?"
This loads the full reference including tools, prompts, resources, configuration options and environment variables, and architecture details.