Featured

Deploy OpenClaw in 60 seconds — 20% off logoDeploy OpenClaw in 60 seconds — 20% off

Launch OpenClaw on Hostinger in about 60 seconds and keep your agent live 24/7. Our referral link gives you 20% off, no coupon code needed.

Launch on Hostinger
Run your Hermes agent on Hostinger, fully managed logoRun your Hermes agent on Hostinger, fully managed

Launch Hermes on Hostinger in one click, fully managed, no VPS knowledge needed. Use code ZACAARON10 for 10% off.

Launch on Hostinger
Crawl and scrape any site into clean data, 10% off logoCrawl and scrape any site into clean data, 10% off

Firecrawl crawls and scrapes any site into clean markdown for your agent. Get 1,000 free credits, and new users get 10% off their first purchase.

Try Firecrawl free
Your own AI agent, running 24/7 with QwikClaw logoYour own AI agent, running 24/7 with QwikClaw

QwikClaw sets up and runs an always-on OpenClaw agent for you. One click, no config files, no server setup.

Deploy now
One API to scrape, enrich, and extract the internet. logoOne API to scrape, enrich, and extract the internet.

Context.dev gives your agents a single API to scrape, enrich, and extract live web data — no proxies, no parsers, no maintenance.

Start building free
SetupClaw: done-for-you OpenClaw for founders & exec teams logoSetupClaw: done-for-you OpenClaw for founders & exec teams

White-glove OpenClaw for founders and exec teams (4–50+ employees): we install, harden, integrate your tools, and maintain it — secured from day one.

Get it set up for you
SEO data APIs for your agent, $1 free credit logoSEO data APIs for your agent, $1 free credit

DataForSEO gives your agent live access to SERP results, keyword data, backlinks, and on-page SEO data through one API. New accounts get a $1 credit, good for up to 20,000 keyword or backlink lookups.

Try DataForSEO free
Reach 47,000+ AI builders

A flat monthly placement in front of developers actively installing AI tools. No lock-in, cancel anytime.

Advertise here

Works with

Claude CodeClaude DesktopCursorVS CodeClineCodex CLIOpenClaw+ any MCP client

Install to Claude Code

This server doesn't publish a one-line install command. Follow the setup in the source repository.

Summary

A TypeScript MCP server delivering a vendored 'dangerous skills' corpus and adversarial fixtures over MCP for security-research testing, supporting SEP-2640 skill delivery with archive-safety hardening.

README.md

dangerous-skills-mcp

A TypeScript MCP server that serves a "dangerous skills" corpus over MCP, implementing the Skills delivery model from SEP-2640. Under an opt-in --adversarial profile it also serves a set of crafted, spec-violating fixtures for testing how MCP hosts handle skill delivery — archive path traversal, decompression bombs, digest/frontmatter mismatches, name collisions, and more.

The corpus is forked from gricha/dangerous-skills (MIT © 2026 Greg Pstrucha). Every payload is benign — it writes a marker file or prints a canary string; nothing performs real harm.

Live endpoint

Deployed as a public Hugging Face Docker Space:

https://olaservo-dangerous-skills-mcp.hf.space/mcp

It's a free CPU Space, so the first request after idle is a slow cold start — retry once. The live Space runs the --adversarial profile. See hf-space/README.md to run or redeploy it.

Any SEP-2640 host can consume it. With fast-agent:

fast-agent go --shell
/mcp connect --name dsk https://olaservo-dangerous-skills-mcp.hf.space/mcp
/skills registry dsk
/skills add check-licenses

Run locally

Requires Node 20+ and pnpm. Runs via tsx (no build step).

pnpm install

# stdio (the default transport)
pnpm serve:stdio                   # faithful corpus only
pnpm serve:stdio -- --adversarial  # + adversarial fixtures

# HTTP (127.0.0.1:3940/mcp by default)
pnpm serve:http

# smoke client — spawns the server and runs conformance checks
pnpm smoke                   # PASS/FAIL per check
pnpm smoke -- --adversarial  # also prints what a conformant host MUST do per fixture

What it serves

Skills are addressed under a skill:// URI scheme:

  • skill://index.json — the catalog (per-skill url + sha256 digest, frontmatter, and archives).
  • skill://<name>/SKILL.md and supporting files — individually addressable and digest-verifiable.
  • skill://<name>.tar.gz / .zip — per-skill archives.

On top of standard MCP resources it adds a resources/directory/read method and advertises the io.modelcontextprotocol/skills capability. The --adversarial profile adds the spec-violating fixtures (namespaced adv-); the smoke client documents each one and the action a conformant host should take. See src/adversarial/catalog.ts for the full list.

Configuration

  • SERVE_PROFILE--adversarial to serve fixtures (the HF image's default), empty for the faithful corpus only.
  • SKILLS_ROOT — corpus root (defaults to the vendored third_party/dangerous-skills/skills).
  • HOST / PORT — HTTP bind address (default 127.0.0.1:3940).
  • ALLOWED_HOSTS, MCP_DISABLE_DNS_REBINDING_PROTECTION — relax the localhost host check for remote hosting behind a proxy.

License

MIT (see LICENSE). The vendored corpus is MIT © 2026 Greg Pstrucha (gricha/dangerous-skills); its notice is kept at third_party/dangerous-skills/LICENSE.

See related servers & alternatives →

Related MCP servers

Browse all →

Related guides

Hand-picked reading to help you choose and use Search servers.