技能 cargo-mcp

📦

cargo-mcp

Name: cargo-mcp
Author: getcargohq

v1.0.3 內容修訂 r1 中風險 ⚙️ 外部命令🌐 網路存取📁 檔案系統存取

Connect Cargo MCP and Run Workspace Actions

Using Cargo through an AI client requires correct authentication, workspace selection, and action inputs. This skill guides hosted MCP connections, cost inspection, batch execution, and result reporting.

支援: Claude Codex Code(CC)

📊 72 尚可

Maintain this Skill? Claim it 📊 72 尚可

使用我的智慧代理安裝

將此請求複製給你的智慧代理，其中包含權威 Skill 頁面和清單。

智能體請求

Review the Skillstore skill "cargo-mcp" from https://skillstore.io/skills/getcargohq-cargo-mcp.md and its manifest at https://skillstore.io/api/skills/getcargohq-cargo-mcp/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

你的智慧代理仍應顯示其計畫，並依安全政策請求必要確認。

Agent 可讀資源

當 AI Agent、爬蟲或腳本需要乾淨脈絡、而不是讀取完整頁面時，請使用這些連結。

Markdown 詳情 GET /skills/getcargohq-cargo-mcp.md 已簽名 manifest GET /api/skills/getcargohq-cargo-mcp/manifest 已簽名 lockfile GET /api/skills/getcargohq-cargo-mcp/lockfile

測試它

正在使用「cargo-mcp」。 Explain how Cargo MCP authenticates and checks my workspace.

預期結果:

OAuth-capable clients discover authentication from the initial 401 challenge.
Headless clients can use a workspace-scoped token supplied through the environment.
Check workspace identity and report its name before performing actions.

正在使用「cargo-mcp」。 Explain whether query_models can join models and calculate totals.

預期結果:

query_models retrieves records from one model using a limit and offset.
It does not support joins, aggregates, or expression filters.
Use the separate Cargo storage CLI workflow for those queries.

正在使用「cargo-mcp」。 What should the summary contain after my approved batch finishes?

預期結果:

Record count: the number processed by the completed batch.
Hit rate: the share producing usable results.
Credits spent: the actual reported charge.
Download: the output CSV link returned by the batch.

安全審計

中風險

v1 • 2026/10/12 開啟版本化報告

All 69 static findings are false positives involving Markdown formatting, documentation links, or intended Cargo connectivity and workspace checks. One semantic concern remains: paid pilot execution precedes the explicit approval step. No evidence found of prompt injection, credential exfiltration, or malicious command execution in the supplied files; hosted service implementation was not reviewed.

已掃描檔案

213

分析行數

複核項目

已忽略的誤報

已確認的安全問題 (1)

中

SKILL.md:172-175

Paid Pilot Before Explicit Approval

The instruction says "Run 10–20 records first" and requests user approval only after reporting the sample and quoting the full batch. An agent following this sequence can spend credits and process personal records before explicit pilot authorization.

The documented sequence explicitly places the pilot before approval of the full run. Prior user authorization could mitigate this, but the skill does not require it.

風險因素

⚙️ 外部命令 (47)

🌐 網路存取 (9)

SKILL.md:3 SKILL.md:5 SKILL.md:6 SKILL.md:18 SKILL.md:24 SKILL.md:65 SKILL.md:72 SKILL.md:82 SKILL.md:97

📁 檔案系統存取 (8)

SKILL.md:32 SKILL.md:42 SKILL.md:43 SKILL.md:44 SKILL.md:45 SKILL.md:177 SKILL.md:182 SKILL.md:191

審計單位: codex

分享與引用此報告

分享版本化評估報告、中性徽章、嵌入卡片與引用。Skillstore 只報告證據，不判斷該 Skill 是否安全。

開啟版本化報告

複製報告連結

https://skillstore.io/skills/getcargohq-cargo-mcp/audits/1?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown 徽章

[![Skillstore security assessment](https://skillstore.io/badges/skills/getcargohq-cargo-mcp/security.svg)](https://skillstore.io/skills/getcargohq-cargo-mcp?utm_source=security_passport_badge)

HTML 徽章

<a href="https://skillstore.io/skills/getcargohq-cargo-mcp?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/getcargohq-cargo-mcp/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

嵌入卡片

<iframe src="https://skillstore.io/embed/skills/getcargohq-cargo-mcp.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

學術引用 (APA · BibTeX · CFF)

APA 引用

getcargohq. (2026). cargo-mcp security audit report (audit version 1) [Author version 1.0.3]. Skillstore. https://skillstore.io/skills/getcargohq-cargo-mcp/audits/1

BibTeX 引用

@techreport{getcargohq-getcargohq-cargo-mcp-2026,
  author = {getcargohq},
  title = {cargo-mcp security audit report (audit version 1)},
  institution = {Skillstore},
  year = {2026},
  number = {1},
  url = {https://skillstore.io/skills/getcargohq-cargo-mcp/audits/1},
  note = {Author version 1.0.3}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "cargo-mcp security audit report (audit version 1)"
version: "1.0.3"
type: report
authors:
  - name: "getcargohq"
date-released: "2026-10-12"
url: "https://skillstore.io/skills/getcargohq-cargo-mcp/audits/1"
identifiers:
  - type: other
    value: "skillstore:getcargohq-cargo-mcp:audit:1"
    description: "Skillstore immutable audit report identifier"

Skillstore 評分

評分依據證據置信度: 中

架構

100

可維護性

內容

社群

規範符合度

你可以打造什麼

Connect an AI Client

Configure hosted Cargo access and verify the correct workspace without installing the Cargo CLI for MCP operations.

Enrich an Approved Batch

Inspect action pricing, obtain pilot approval, test a small sample, and submit an approved batch with a downloadable result.

Configure Headless Access

Connect a headless client with an environment-supplied workspace token and inspect available tools, schemas, and recent usage.

試試這些提示

Connect Cargo

Guide me through connecting Cargo MCP to Claude Desktop with OAuth. Verify and report the workspace before running any action.

Find and Price an Action

Find available company enrichment actions in my workspace. Show their credit costs and required inputs. Do not execute any action.

Plan a Controlled Batch

Plan enrichment for my approved records. Request pilot budget approval before execution. Report sample results, then request separate approval for the remaining batch.

Prepare Headless Automation

Plan headless Cargo MCP access using an environment token. Verify workspace identity, discover tools, inspect schemas, and define polling without exposing credentials.

最佳實務

Verify and report the authenticated workspace before reading data or running actions.
Approve pilot costs and data processing before execution, then approve the full batch estimate separately.
Inspect action schemas, preserve returned action descriptors, and supply inputs through data or records.

避免

Running paid pilot records without explicit user authorization.
Placing action inputs in config or looping single-record calls instead of using batch execution.
Treating a limited model query as complete SQL analysis or counting truncated results.

常見問題

Does hosted Cargo MCP require the Cargo CLI?

Hosted MCP operations do not require the Cargo CLI. Separate platform administration tasks use CLI skills; some installer metadata also declares the CLI.

Which AI tools can use this skill?

The skill declares Claude, Codex, and Claude Code support. Actual execution requires a client configured for the hosted Cargo MCP server.

Why does the first connection return 401?

The initial unauthenticated response provides OAuth discovery information. Continued failures after authentication require checking client configuration and credentials.

How should I handle an API token?

Use a workspace-scoped token from Cargo Settings and supply it through the environment. Do not paste its value into configuration or reports.

Can actions spend credits or process personal data?

Yes. Approve pilot and batch budgets before execution, and confirm a lawful basis, suppression checks, and relevance before processing personal records.

Why do tools differ between workspaces?

The endpoint includes platform tools and workspace-published tools. Inspect the connected tool list and verify the workspace before choosing an action.