CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

26.3 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 1 flagged

Securityscore 75 · 1 finding

HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · package.json

HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.

Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted

excerptpackage.json· json

78]

79},

80"packageManager": "[email protected]+sha512.664074abc367d2c9324fdc18037097ce0a8f126034160f709928e

… (76 chars elided on L80)

81}

Occurrences

1 occurrence · at L80

How to fix

Remove the encoded blob, or decode it and review what it actually contains.

Decode the base64 string and confirm it is not an instruction directed at the agent.
Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha256a33ffe00d5643d30…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

README.md · 1 file

README.md6.3 KB · Markdown

Hound MCP

The dependency bloodhound for AI coding agents.

Hound MCP Deployment Diagram

Why Hound?

AI coding agents recommend and install packages without knowing if they're safe — and most security tools require accounts, API keys, or paid plans to tell you. Hound fixes that: it scans for vulnerabilities, checks licenses, audits dependency trees, and detects typosquatting across 7 ecosystems — zero config, zero API keys, zero cost.

Hound is the only security tool built specifically for AI coding agents — works across npm, PyPI, Go, Cargo, Maven, NuGet, and RubyGems, and plugs into Claude Code, Cursor, VS Code, and any MCP client out of the box.

It uses two fully free, unauthenticated public APIs: [deps.dev](https://deps.dev) (Google Open Source Insights) and [OSV](https://osv.dev) (Google Open Source Vulnerabilities).

Quickstart

Claude Code

claude mcp add hound -- npx -y hound-mcp

Claude Desktop / Cursor / Windsurf

Add to your MCP config file:

{
  "mcpServers": {
    "hound": {
      "command": "npx",
      "args": ["-y", "hound-mcp"]
    }
  }
}

VS Code (Copilot)

{
  "mcp": {
    "servers": {
      "hound": {
        "type": "stdio",
        "command": "npx",
        "args": ["-y", "hound-mcp"]
      }
    }
  }
}

#### Config file locations

Client	Config path
Claude Desktop (macOS)	`~/Library/Application Support/Claude/claude_desktop_config.json`
Cursor	`~/.cursor/mcp.json`
Windsurf	`~/.codeium/windsurf/mcp_config.json`

Tools

12 tools → Full reference with example outputs

Tool	What it does
`hound_audit` ⭐	Scan an entire lockfile for vulnerabilities across all dependencies
`hound_score`	0–100 Hound Score (vulns + scorecard + recency + license) with letter grade
`hound_compare`	Side-by-side comparison of two packages with a recommendation
`hound_preinstall`	GO / CAUTION / NO-GO verdict before installing a package
`hound_upgrade`	Find the minimum safe version upgrade that resolves all known vulns
`hound_license_check`	Scan a lockfile for license compliance against a policy
`hound_vulns`	All known vulnerabilities for a package version, grouped by severity
`hound_inspect`	Full package profile — license, vulns, scorecard, stars, dep count
`hound_tree`	Full resolved dependency tree with transitive deps
`hound_typosquat`	Detect typosquatting variants of a package name
`hound_advisories`	Full advisory details by GHSA, CVE, or OSV ID
`hound_popular`	Scan popular packages for known vulnerabilities

Supported ecosystems: npm · pypi · go · maven · cargo · nuget · rubygems

Built-in Prompts

3 prompts you can invoke directly from your AI client. → Full prompt reference

Prompt	What it does
`security_audit`	Full project security audit — vulns, licenses, typosquats
`package_evaluation`	Go/no-go recommendation before adding a new dependency
`pre_release_check`	Pre-ship dependency scan that flags release blockers

Use Cases

→ See full examples with real lockfiles and expected output

Before merging a PR — scan the lockfile diff to catch newly introduced vulnerabilities before they land in main
Auditing an inherited codebase — run hound_audit on an existing lockfile to get a full report in seconds
Checking a package before adding it — use hound_preinstall to get a GO / CAUTION / NO-GO verdict
License compliance — run hound_license_check to ensure no GPL or AGPL packages sneak into a commercial project
CI security gate — ask your AI agent to run a security audit as part of every release check

Local Development

git clone https://github.com/tiluckdave/hound-mcp.git
cd hound-mcp
pnpm install
pnpm build
pnpm test         # run tests
pnpm check        # typecheck + lint + test

Roadmap

[ ] Docker support — run Hound as a container for CI/CD pipelines
[ ] `bun.lockb` parser — Bun lockfile support
[ ] `gradle.lockfile` parser — Gradle (Java/Android) ecosystem support
[ ] `hound_diff` tool — compare two lockfile snapshots to surface newly introduced risks
[ ] GitHub Action — run hound_audit as a PR check without an AI agent

Contributing

Contributions are welcome. Read CONTRIBUTING.md to get started.

The one rule: Hound must stay zero-config and free forever. Don't add features that require API keys or accounts.

Good first issues are labeled and ready.

Community

💬 Questions or ideas? Open a Discussion

Hound Mcp— agent skill

Is Hound Mcp safe to install?

Hound MCP

Why Hound?

Quickstart

Claude Code

Claude Desktop / Cursor / Windsurf

VS Code (Copilot)

Tools

Built-in Prompts

Use Cases

Local Development

Roadmap

Contributing

Community

License

Audit the pieces. Scan the whole. Decide.