{
"markdown": "# dsh-skills-anywhere\n\n**Your skills, anywhere.** Install an [Agent Skill](https://agentskills.io) once, use it in every agent: a live skill provider for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh`) and an MCP server for Claude Code, Cursor, Codex and friends.\n\nEnglish | [中文](README.zh.md)\n\n**[Try the interactive Hugging Face playground](https://huggingface.co/spaces/glayguo/dsh-skills-anywhere)** — explore an example workspace, resolve name clashes, and search beyond the catalog budget. No installation or model API needed. [How it works](docs/HUGGINGFACE.md).\n\n**Bring your own `SKILL.md`.** Compare the provider's strict and lenient parsing\nlocally: inspect repairs, invocation settings and a downloadable check report.\nYour file stays in the browser. The same checks are available in the\n[command line and CI](docs/CHECKING.md), with file hashes and actionable exit codes.\nThis is a parser check, not a security audit or a guarantee of compatibility with every client.\n\n\n\n[](https://github.com/noteflowai/dsh-skills-anywhere/actions/workflows/ci.yml)\n[](https://www.npmjs.com/package/dsh-skills-anywhere)\n[](https://github.com/topics/dsh-plugin)\n[](https://scorecard.dev/viewer/?uri=github.com/noteflowai/dsh-skills-anywhere)\n[](LICENSE)\n\nAgent Skills are portable by design: a folder with a `SKILL.md`. Every agent still looks only in its own folder, so a skill you install for Claude Code is invisible to Codex, Cursor and DeepSeek Harness, and the ones you wrote for them are invisible back. `dsh-skills-anywhere` reads all of those folders where they live and serves them everywhere: as a live skill provider inside dsh, and as an MCP server for Claude Code, Cursor, Codex and any other MCP client.\n\n
\n\nInside dsh, it registers one extra provider on the built-in `ctx.skills` registry, so the model's normal `skill` tool and `/name` invocation simply see more skills:\n\n- **Every other agent's skill directories.** 60+ agents out of the box: Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Kiro, Goose, OpenCode, Roo, Cline, Qwen Code, Trae and more. Project-level and user-level.\n- **Claude Code plugin marketplaces.** The skills nested inside `~/.claude/plugins/marketplaces/*/plugins/*/skills/*`, including the official Anthropic marketplace.\n- **Any git repository full of skills.** Point at `anthropics/skills`, a sub-directory, a branch, a tag, or a commit. It is shallow-cloned into a local cache, refreshed in the background, and pinned in a lock file.\n- **Zero copies, zero symlinks.** Files are read where they live and re-read on every load. Edit a skill in Cursor and dsh sees the change. Nothing to import, nothing to keep in sync.\n\nHundreds of skills would bloat every model request, so the provider keeps a **catalog budget**: at most 50 skills enter the model's session catalog by default, and the rest stay one `find_skills` call away through two small tools the plugin adds, with `/name` invocation untouched.\n\nThe same pool is available **outside dsh** too: `dsh-skills-anywhere mcp` serves it to any [MCP](https://modelcontextprotocol.io) client (Claude Code, Cursor, Codex, Windsurf…) as `find_skills` / `open_skill` tools and `skill://` resources, so one install of a skill reaches every agent you use.\n\nIt also **deduplicates** symlinked and byte-identical installs (the `skills` CLI links one canonical copy into several agents), **repairs** common frontmatter drift instead of silently dropping a skill, and **renames** colliding names (`discord/configure` vs `telegram/configure`) so every skill stays reachable. A small CLI shows you exactly what dsh will see and why.\n\n## Quick start\n\n```sh\n# 1. Install into the dsh profile you use (web is the default UI profile)\ndsh plugin --profile web add dsh-skills-anywhere\n\n# 2. See what the model will get, without booting dsh\nnpx dsh-skills-anywhere list\n\n# 3. Add a whole repository of skills\nnpx dsh-skills-anywhere add anthropics/skills\n```\n\nPublished on [npm](https://www.npmjs.com/package/dsh-skills-anywhere) with build provenance; every [GitHub release](https://github.com/noteflowai/dsh-skills-anywhere/releases) also carries the same tarball.\n\nStart dsh as usual. The skill catalog now includes everything above; load a skill with the `skill` tool or `/skill-name` exactly as before.\n\nOn a machine with only Claude Code installed, `list` already finds the 31 skills inside the official plugin marketplace, none of which dsh sees on its own:\n\n```\n$ npx dsh-skills-anywhere list\nNAME FROM PATH\ndiscord-access claude plugin discord @ claude-plugins-official ~/.claude/plugins/marketplaces/.../discord/skills/access/SKILL.md\nfrontend-design claude plugin frontend-design @ claude-plugins-official ~/.claude/plugins/marketplaces/.../frontend-design/skills/frontend-design/SKILL.md\nskill-creator claude plugin skill-creator @ claude-plugins-official ~/.claude/plugins/marketplaces/.../skill-creator/skills/skill-creator/SKILL.md\n...\n31 skills, 6 renamed — run `dsh-skills-anywhere doctor` for details\n```\n\n\nInstall from a git checkout or a release tarball instead of npm\n\nEvery [GitHub release](https://github.com/noteflowai/dsh-skills-anywhere/releases) carries a prebuilt tarball, and both `dsh plugin add` and `npx` accept its URL directly (`https://github.com/noteflowai/dsh-skills-anywhere/releases/download/v0.6.0/dsh-skills-anywhere-0.6.0.tgz`). If you want an unreleased commit:\n\n```sh\ndsh plugin --profile web add github:noteflowai/dsh-skills-anywhere\n```\n\nA git install ships sources, so pnpm has to run this package's `prepare` build. pnpm 10+ refuses until you allow it: the first `add` fails and prints the exact key to allow. Copy that key (it includes the commit) into the profile's `pnpm-workspace.yaml` and run the `add` again.\n\n```yaml\n# $DSH_HOME/profiles/web/pnpm-workspace.yaml\nallowBuilds:\n 'dsh-skills-anywhere@https://codeload.github.com/noteflowai/dsh-skills-anywhere/tar.gz/': true\n```\n\nPin a commit (`github:noteflowai/dsh-skills-anywhere#`) if you want the install to be reproducible.\n\n\n\n\nRequirements\n\n- DeepSeek Harness `0.1.5-rc.1` or newer (the suite runs against `0.1.5-rc.1` and `0.1.5-rc.2`), any profile that mounts `@deepseek-ai/dsh-skill` (the shipped `web`, `acp`, `headless` and `sdk` profiles all do)\n- Node.js 22.19+ or 24+\n- `git` on `PATH` for git sources (everything else works without it)\n\n\n\n## What gets discovered\n\n| Where | Example | dsh source label | Default rank |\n|---|---|---|---|\n| Another agent's **project** skills | `/.claude/skills/*` | `anywhere-project` | 250 |\n| Another agent's **user** skills | `~/.codex/skills/*`, `~/.cursor/skills/*` | `anywhere-user` | 550 |\n| **Claude Code plugin marketplaces** and the installed-plugin cache | `~/.claude/plugins/marketplaces/*/plugins/*/skills/*` | `anywhere-claude-plugins` | 580 |\n| **Git sources** | `anthropics/skills`, `vercel-labs/agent-skills/skills` | `anywhere-source` | 700 |\n\nLower rank wins a duplicate name inside the dsh registry. The built-in dsh roots keep their ranks (`.dsh/skills` 100, `.agents/skills` 200, `~/.dsh/skills` 400, `~/.agents/skills` 500), so a skill you wrote for dsh always beats the same name found elsewhere. `.agents/skills` and `.dsh/skills` are deliberately not re-scanned here.\n\nRun `npx dsh-skills-anywhere agents` for the full agent table and which directories exist on your machine.\n\n### Skill format\n\nAny directory with a `SKILL.md` following the [Agent Skills specification](https://agentskills.io/specification), plus dsh's flat `.md` form. `name`, `description`, `license`, `compatibility`, `allowed-tools`, `metadata`, and dsh's `disable-model-invocation` / `user-invocable` are all understood. Unknown frontmatter (Claude Code's `argument-hint`, `context`, ...) is preserved under `metadata.frontmatter`. `scripts/`, `references/` and `assets/` are exposed through the skill's resource directory like any dsh skill.\n\nIn the default **lenient** mode a missing name falls back to the directory, an invalid name is normalised to kebab-case, and a missing description is derived from the first paragraph. Each repair is recorded and shown by `doctor`. Set `lenient: false` to match the strict behaviour of the built-in provider.\n\n## Git sources\n\n```sh\nnpx dsh-skills-anywhere add anthropics/skills # default branch\nnpx dsh-skills-anywhere add anthropics/skills@v1.0.0 # tag or branch\nnpx dsh-skills-anywhere add vercel-labs/agent-skills/skills # sub-directory\nnpx dsh-skills-anywhere add https://github.com/o/r/tree/main/dir # GitHub tree URL\nnpx dsh-skills-anywhere add git@gitlab.com:group/skills.git # any git URL\nnpx dsh-skills-anywhere add ./local/skills-repo --project # local repo, project-scoped\nnpx dsh-skills-anywhere add o/r --ref 3f2a9c1 --rank 300 # pin a commit, set precedence\n```\n\nSources come from three places, merged in this order: the plugin `config.sources`, the user file `~/.dsh/skills-anywhere/sources.json`, and the project file `/.dsh/skills-anywhere.json` (commit it to share skills with your team). The CLI edits the last two.\n\nEach repository is shallow-cloned once into `~/.dsh/skills-anywhere/cache///` (`@` when a branch, tag or commit is set, so several refs of one repository never share a checkout) and refreshed when dsh starts, every `syncIntervalMs` (6 hours by default), and whenever a sources file changes. The resolved commit of every source is written to `~/.dsh/skills-anywhere/lock.json`. Discovery only ever reads the cache, so a failed refresh means yesterday's skills, never an empty catalog. The catalog is invalidated as soon as a refresh brings changes; dsh never waits on the network.\n\n## Catalog budget and the `find_skills` / `open_skill` tools\n\ndsh publishes every model-invocable skill's name and description into the session, on every request. With marketplaces and a few git sources that is hundreds of lines of context. The provider therefore ranks its skills and marks only the first `catalog.limit` (default 50) as model-invocable; the remainder is published with model invocation off, which keeps it out of the catalog but still loadable by you with `/name`.\n\nTwo tools, registered by the `dsh-skills-anywhere/tools` row, make the hidden part reachable for the model:\n\n- **`find_skills(query, limit?)`** searches every skill by keyword (name, description, `whenToUse`, origin), catalog or not, and says which matches are listed.\n- **`open_skill(name)`** loads any skill by exact name, including ones the budget hid. Skills whose own frontmatter says `disable-model-invocation: true` are still refused, exactly as the built-in `skill` tool does.\n\n```yaml\n- id: skills-anywhere\n config:\n catalog:\n limit: 30 # 0 = unlimited (old behaviour)\n pin: [frontend-design] # always listed\n hide: [example-skill] # never listed, still searchable and /name-invocable\n- id: skills-anywhere-tools\n config:\n findLimit: 10\n```\n\nAuthor-disabled skills never count against the budget. Which skills stay listed follows the precedence order below, so project-level skills win over user-level, which win over marketplaces and git sources. The tools row needs the tool runtime (`ctx.tools`); in a profile without one it stays pending and the provider works alone.\n\n## CLI\n\n**Check before committing.** Run `npx -y dsh-skills-anywhere@0.6.0 check\nskills/example/SKILL.md --fail-on-repair`. The same parser used in the playground\nnow has batch file checks, JSON reports with file hashes, and CI exit codes.\nChecks read only the named files. [Commands, CI example and scope](docs/CHECKING.md).\n\n```\ndsh-skills-anywhere list [--all] [--json] Skills the provider publishes (--all shows hidden duplicates)\ndsh-skills-anywhere agents [--json] Supported agents and which directories exist here\ndsh-skills-anywhere sources [--json] Configured git sources and their synced commits\ndsh-skills-anywhere add [--ref] [--path] [--rank] [--project]\ndsh-skills-anywhere remove [--project]\ndsh-skills-anywhere sync [--force] [--json] Clone or refresh every source now\ndsh-skills-anywhere doctor [--json] Repaired, skipped, renamed and duplicate skills, with reasons\ndsh-skills-anywhere check [--json] Explicit local files; strict parser gate by default\ndsh-skills-anywhere mcp Serve the same skills to any MCP client over stdio\n```\n\nAll commands accept `--cwd `. For `check`, it resolves the named files; other\ncommands use it to pick the project. `check --lenient` accepts provider repairs;\n`--fail-on-repair` rejects any reported repair in the selected mode.\nThe CLI uses the same parsing code as the plugin and never needs dsh running.\n\n## Use as an MCP server\n\nSkills are not a dsh-only idea, and neither is this provider. `dsh-skills-anywhere mcp` starts a [Model Context Protocol](https://modelcontextprotocol.io) server over stdio that exposes the identical pool (agent directories, Claude Code marketplaces, git sources, same dedupe and rename rules) to any MCP client:\n\n| Tool | What it does |\n| --- | --- |\n| `list_skills` | Browse every model-invocable skill with its description and origin (`limit`, `offset`) |\n| `find_skills` | Keyword search across names, descriptions and origins |\n| `open_skill` | Load one skill's instructions plus the directory its scripts and references live in |\n\nSkills are also exposed as `skill://` resources (with completion), for clients that let you @-mention resources. Skills whose frontmatter sets `disable-model-invocation: true` are never listed or opened. The server needs no dsh installation at all.\n\n**Claude Code** (as a plugin; this repo doubles as a plugin marketplace)\n\n```sh\nclaude plugin marketplace add noteflowai/dsh-skills-anywhere\nclaude plugin install dsh-skills-anywhere@noteflowai\n```\n\nOr register the bare server instead: `claude mcp add skills-anywhere -- npx -y dsh-skills-anywhere mcp`. Either way, restart Claude Code once so it connects.\n\n**Cursor** (`.cursor/mcp.json` or `~/.cursor/mcp.json`)\n\n```json\n{ \"mcpServers\": { \"skills-anywhere\": { \"command\": \"npx\", \"args\": [\"-y\", \"dsh-skills-anywhere\", \"mcp\"] } } }\n```\n\n**Codex** (`~/.codex/config.toml`)\n\n```toml\n[mcp_servers.skills-anywhere]\ncommand = \"npx\"\nargs = [\"-y\", \"dsh-skills-anywhere\", \"mcp\"]\n```\n\nThe server is also listed in the [official MCP registry](https://registry.modelcontextprotocol.io) as `io.github.noteflowai/dsh-skills-anywhere`, so registry-aware clients can install it by name. The repository is also an [Agent Plugin](https://agent-plugins.org) (`plugin.json` + `mcp.json` at the root), so Cursor and other open-plugin clients can install it from the repository URL. Add `--cwd ` when the client does not start the server inside the project you are working on. Git sources sync in the background on start, exactly as in dsh. Programmatic use: `import { createSkillsAnywhereServer } from 'dsh-skills-anywhere/mcp'` returns the `McpServer` and the provider so you can attach your own transport.\n\n## Browse and toggle skills in the dsh web UI\n\nIn `dsh web`, open **Settings → Plugins → Plugin configuration**. The **Skills Anywhere** card lists every skill the provider found, grouped by where it lives (agent directories, Claude Code plugins, git sources), with its catalog state — *listed* for the model, *not listed* (kept out by the budget or by you) or *author disabled* — and the name it was renamed to when it collided. Each row offers **Pin** (always listed), **Hide** (out of the model catalog, still `/name`- and `find_skills`-reachable) and **Exclude** (dropped from the provider); the catalog budget is editable in place, and a filter box searches names, descriptions and origins.\n\n
![\"Actual](\"docs/skill-check.png\")
![\"dsh-skills-anywhere](\"docs/demo.gif\")
![\"The](\"docs/web-card.png\")