{ "markdown": "

Language: English · 中文

\n\n

\n \"yotta-verify-mcp\n

\n\n

yotta-verify-mcp · 元信MCP (YuanXin MCP)

\n\n

YottaMeta's pre-install security scanner for Agent skills, exposed as a\nstdio MCP server. Before you install any skill, plugin or MCP server, it runs a\ndeterministic static scan — prompt injection, malicious patterns, SKILL.md integrity and\npermissions — and returns a verdict, an audited badge, a CI gate and a\nreport as MCP tools.

\n

Activates when configuring the 元信 MCP server in an MCP client, wiring a\ntrust-scan into an agent / workflow, or calling the MCP tools.

\n

Zero dependencies (Python 3.8+ standard library); Windows + Linux + macOS; fully\nlocal and offline — no network calls, no execution of the scanned code.

\n\n

\n \"License:\n \"Standard:\n \"npm\n \"GitHub\n \"last\n \"PRs\n

\n\n## What it is\n\nThe skill / plugin market has a trust problem: a 2025 survey of 22,511 skills found 140,963 issues,\nand **36% contain prompt injection**. 元信 MCP gives you a deterministic answer before you\ninstall — the same scan as the [yotta-verify](https://github.com/YottaMeta/yotta-verify) CLI,\nexposed as four MCP tools so any MCP client (Claude, VS Code, Codex, Cursor, …) can call it.\n\nIt is a pre-install verifier, not a sandbox and not a runtime monitor: it only reads files\nand prints a report. It never executes the scanned code, never connects to the network for the scan,\nand never fixes anything.\n\n## Why use it\n\n| Advantage | Description |\n|---|---|\n| **Trust before install** | A deterministic verdict for any skill / MCP server, instead of \"trust me\" |\n| **Zero dependency** | Python 3.8+ standard library; no daemon / database / network |\n| **Fully local offline** | Scans directories and npm tarballs on disk; nothing is executed or uploaded |\n| **Drop into any MCP client** | Standard stdio MCP server — configure the server, and the four tools appear |\n| **Family synergy** | Same rules table as yotta-verify (single source); verdicts merge with yotta-vetter / yotta-security-audit |\n| **Free & open** | MIT; the whole scanner is free |\n\n## MCP tools\n\n| Tool | What it does |\n|---|---|\n| `scan_skill` | Pre-install scan: `target` (dir / .tgz / npm package) → verdict + severity counts + findings |\n| `generate_badge` | Audited badge: local SVG + shields.io URL; folds in validate / vetter / audit / version / tests |\n| `gate_check` | CI gate: fail when the worst severity exceeds `max_severity` (default medium) |\n| `get_report` | Verification report: Markdown or JSON, same format as the CLI |\n\n## MCP client configuration\n\nYou usually do not need to write the `mcpServers` entry yourself: after installing this skill, an AI agent auto-adds the `yotta-verify-mcp` entry per the「AI 自动接入」section in `SKILL.md`, and falls back to the CLI scanner when MCP tools are unavailable.\n\n## Tool reference\n\n### `scan_skill`\n\nScan a skill directory or package before install.\n\n| Param | Type | Required | Meaning |\n|---|---|---|---|\n| `target` | string | yes | Skill directory path, `.tgz` / `.tar.gz` path, or npm package name (auto `npm pack` to a temp dir, then scan) |\n\nReturns a JSON result: verdict, severity counts, and findings (prompt injection / malicious\npatterns / SKILL.md integrity).\n\n### `generate_badge`\n\nGenerate an audited badge (local SVG + shields.io URL).\n\n| Param | Type | Meaning |\n|---|---|---|\n| `target` | string | Optional: scan this to derive the verdict |\n| `verdict` | string | Optional: set the verdict directly |\n| `validate` | string | Optional: `pass` / `fail` (validate-skill result) |\n| `vetter` / `audit` | string | Optional: external verdicts to fold in |\n| `version` | string | Optional: version label. Defaults to the scanner (yotta-verify) version (e.g. 0.1.1) |\n| `tests` | integer | Optional: engine test count |\n| `out` | string | Optional: write the SVG to this path |\n\n> Note: the badge's `version` segment reflects the version of the scanning engine\n> (yotta-verify), not the MCP package (0.1.4). Pass `version` to override.\n\n### `gate_check`\n\nCI pre-install gate.\n\n| Param | Type | Meaning |\n|---|---|---|\n| `target` | string | Required: dir / package to scan |\n| `max_severity` | string | Optional: `info` / `low` / `medium` / `high` / `critical` (default `medium`) |\n\nReturns `pass`, `verdict`, `worst`, `max_severity` and an exit `code`.\n\n### `get_report`\n\nGenerate a verification report.\n\n| Param | Type | Meaning |\n|---|---|---|\n| `target` | string | Required: dir / package to scan |\n| `format` | string | Optional: `json` / `markdown` (default `markdown`) |\n| `out` | string | Optional: write the report to this path |\n\n## Boundary\n\nThis is a **local, offline, static** scan:\n\n- **Directory scan** is fully offline — content never leaves your machine.\n- **npm package scan** only downloads the public package into a temporary directory (then removes it);\n it does not upload your content and does not execute the scanned package code.\n- It does **not** perform dynamic analysis, does **not** fix anything, and does **not** make the final\n decision. Treat the verdict as a strong signal and confirm any \"install / don't install\" decision\n yourself.\n- Only scan targets you are authorised to evaluate.\n\n## Installation of the skill\n\nThe package also ships a `SKILL.md` so an agent can learn how to configure and use the MCP server.\nPick any of the four methods below (skill files come from **npm**; GitHub can be slow without a proxy).\n\n### Method 1: npm one-liner (recommended)\n\n```text\n# Optional China mirror: npm config set registry https://registry.npmmirror.com\nnpx -y @yottameta/yotta-verify-mcp --agent # install to the agent's default user-level skills dir\nnpx -y @yottameta/yotta-verify-mcp --dir # point to the skills dir itself (e.g. ~/.codex/skills)\n```\n\n- `--agent ` installs to that agent's default user-level directory; `--list` shows each agent's default directory.\n- `--dir ` installs to the given directory.\n- The installer also starts an MCP server when run with no arguments: `npx -y @yottameta/yotta-verify-mcp`.\n\n### Method 2: git clone (developers / git available)\n\n```text\ngit clone https://github.com/YottaMeta/yotta-verify-mcp.git /yotta-verify-mcp\n```\n\n### Method 3: GitHub Download ZIP (manual / no git)\n\nOn the GitHub repository `YottaMeta/yotta-verify-mcp`, click **Code → Download ZIP**, unzip it and put\nthe `yotta-verify-mcp` folder into the agent's skills directory.\n\n### Method 4: install.sh (multi-agent one-liner script)\n\n```text\nbash install.sh --agent # install to the agent's default user-level directory\nbash install.sh --dir # install to the given directory\nbash install.sh --list # list agents -> default directories\n```\n\n## Development & validation\n\nThe package ships its own test suite (included in the published package):\n\n```bash\n# Run the full suite (32 cases) from the skill directory (Python 3.8 / 3.13 both green)\npython scripts/test_yotta_verify_mcp.py\n\n# Run the MCP server directly for debugging\npython scripts/yotta_verify_mcp.py\n```\n\nReferences: `references/trust-checklist.md` (pre-install trust checklist for MCP servers / plugins).\n\n## License\n\nMIT © YottaMeta — see [LICENSE](./LICENSE).\n", "bytes": 8336, "sha": "e36d5d0e787f595f66964bc264b06cdaf38a8e4cd741ceebc3cd3a9ef0b64f42", "repo_slug": "yottameta/yotta-verify-mcp", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/mcp_io_github_yottameta_yotta_verify_mcp_676311f3/readme" }