Back to skill

Security audit

Cloudways MCP

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious with medium confidence: scanners report an undisclosed Slack reporting destination, unsafe automation examples, and incomplete inspection of two referenced files.

Review the skill's reporting and automation instructions before installing, especially where reports are sent, when webhook-triggered backups run, and how credentials are entered in commands. The scan is incomplete: two referenced files were only partially inspected, so the results do not establish that the remaining content is safe. The verdict is suspicious, not a finding that the author acted maliciously.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (78)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The Cloudways monitoring and report generation align with the declared purpose. However, the script also explicitly sends the resulting operational report outside Cloudways via an HTTP POST. This external disclosure is a distinct, undeclared capability rather than merely an implementation detail of MCP monitoring. The destination is environment-configured, so the code does not establish that it is attacker-controlled. The script requests no Cloudways write operations and explicitly instructs Claude to avoid credential-returning tools.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

md
is the **official article**: `support.cloudways.com/en/articles/14654372`. See `references/installation.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 27)May include surrounding context.

md
| Initial installation/configuration of the MCP server | `references/installation.md` |

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 33)May include surrounding context.

md
| Multiple Cloudways accounts / multi-account configuration | `references/installation.md` (Multi-account section) |

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 142)May include surrounding context.

md
- `X-Mcp-Host` — the client identifier (`claude-code` / `claude-desktop`; full list of official values in `references/installation.md`)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 146)May include surrounding context.

md
migrated such connections retain unrestricted full-account access. Migration: `references/installation.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 150)May include surrounding context.

md
For the full connection, role guidance, and multi-account setup, see `references/installation.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 164)May include surrounding context.

md
— one connection per account, each with its own credentials) is documented in `references/installation.md` section **Multi-account configuration**. The runtime

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 30)May include surrounding context.

md
| Cache clear, SSL, backup, restart, IP whitelist | `references/workflows-maintenance.md` |

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 51)May include surrounding context.

md
5. **`server_delete`, `app_delete`, and `app_cname_delete` = immediate destruction in production.** Requires double confirmation (W!): of both the operation and the specific domain/application/server.

6. **Credentials.** Each account authenticates with its own **Access Token** (case-sensitive `X-Access-Token` header; roles + legacy-key migration in the Authentication section below). Don't print tokens in responses. Don't mix credentials between accounts. If the user asks to see them, refer them to platform.cloudways.com → API section.

7. **Don't sweep the credential-returning tools.** `server_get`, `app_get` and `app_credentials` return master, database and SSH credentials inside their ordinary payloads. Running one of them over **every** server or **every** app pulls the account's secrets into the conversation, where they stay for the rest of it — and telling yourself to keep them out of the *report* comes too late to help. Use `server_list` / `app_list` for inventory; call the oth

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 141)May include surrounding context.

md
The official MCP is hosted at `https://mcp.cloudways.com/mcp/` and authenticates via two **case-sensitive** HTTP headers:

- `X-Access-Token` — a Cloudways **Access Token** generated in the platform
- `X-Mcp-Host` — the client identifier (`claude-code` / `claude-desktop`; full list of official values in `references/installation.md`)

Tokens are **role-based (RBAC)**: **READ** (look-ups only), **LIMITED** (selected endpoint groups), **FULL ACCESS** (everything, including destructive actions). Start new integrations with READ and escalate only when a connection must write. This platform-level control **complements** (does not replace) this skill's write-confirmation discipline — even FULL ACCESS has no per-tool gating at the MCP layer.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 178)May include surrounding context.

md
- **IDs don't cross accounts.** A server_id / app_id you received from `mcp__cloudways-clientA` is valid **only** against clientA. Never take an ID from one account's response and pass it to a tool of another connection.
- **Per-account confirmation.** A write confirmation on one account does not apply to another. Every write operation on a new account = a new confirmation block (including the account line).
- **Credentials don't mix.** Each connection has its own Access Token. Don't assume the same credentials work on another account.

### Cross-account search (read only)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 15)May include surrounding context.

md
---

## Step 1 — Generate an Access Token

1. Log in to [platform.cloudways.com](https://platform.cloudways.com).
2. Open the **API** section (bottom-left of the platform — the same place the legacy API key lived).

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 19)May include surrounding context.

md
1. Log in to [platform.cloudways.com](https://platform.cloudways.com).
2. Open the **API** section (bottom-left of the platform — the same place the legacy API key lived).
3. Generate an **Access Token** (Access Token Details → Create Access Token). Note: only the **primary account owner** can create/manage API credentials — team-member accounts have no API Integration section. Give the token a name, an expiration period (1 day → never; shortest that works), and a **role**:
   - **READ** — look-ups only (status, config, monitoring). Recommended starting point for every new integration, and for monitoring-only connections.
   - **LIMITED** — only the endpoint groups you select.
   - **FULL ACCESS** — everything the account can do, including destructive actions. Only for connections that genuinely need to make changes.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 19)May include surrounding context.

md
1. Log in to [platform.cloudways.com](https://platform.cloudways.com).
2. Open the **API** section (bottom-left of the platform — the same place the legacy API key lived).
3. Generate an **Access Token** (Access Token Details → Create Access Token). Note: only the **primary account owner** can create/manage API credentials — team-member accounts have no API Integration section. Give the token a name, an expiration period (1 day → never; shortest that works), and a **role**:
   - **READ** — look-ups only (status, config, monitoring). Recommended starting point for every new integration, and for monitoring-only connections.
   - **LIMITED** — only the endpoint groups you select.
   - **FULL ACCESS** — everything the account can do, including destructive actions. Only for connections that genuinely need to make changes.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 19)May include surrounding context.

md
1. Log in to [platform.cloudways.com](https://platform.cloudways.com).
2. Open the **API** section (bottom-left of the platform — the same place the legacy API key lived).
3. Generate an **Access Token** (Access Token Details → Create Access Token). Note: only the **primary account owner** can create/manage API credentials — team-member accounts have no API Integration section. Give the token a name, an expiration period (1 day → never; shortest that works), and a **role**:
   - **READ** — look-ups only (status, config, monitoring). Recommended starting point for every new integration, and for monitoring-only connections.
   - **LIMITED** — only the endpoint groups you select.
   - **FULL ACCESS** — everything the account can do, including destructive actions. Only for connections that genuinely need to make changes.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 27)May include surrounding context.

md
> Treat the token like a password — never commit it or print it in responses. Prefer one token **per integration** (per MCP connection), each with the minimum role it needs, so tokens can be revoked individually.

> **Legacy API key — deprecated.** The old flow (API key + `X-CW-Email`/`X-CW-Api-Key` headers) still works but the API key **stops working on October 15, 2026**. If you have an existing connection using the old headers, regenerate as an Access Token and update the connection before then. Until migrated, a legacy connection retains **unrestricted full-account access** — the RBAC roles apply only to Access Tokens. And note there is still **no per-tool permission control at the MCP layer** beyond the token's role: a FULL ACCESS token can call every tool.

**Required headers** (every request; header names are case-sensitive):

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 33)May include surrounding context.

md
| Header | Value |
|--------|-------|
| `X-Access-Token` | your Cloudways Access Token |
| `X-Mcp-Host` | the client you connect from — official values: `claude-code`, `claude-desktop`, `cursor`, `Devin` (the client formerly called Windsurf — the article's own snippet sends the capitalised `Devin`, and header values are case-sensitive), `vs-code`, `gemini-cli`, `codex`, `codex-cli` |

---

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 259)May include surrounding context.

md
> a collapsed fallback. `npx mcp-remote@0.14.0` pins the named package and **nothing underneath
> it**: the ~80 transitive dependencies are resolved fresh whenever the npx cache is empty, so a
> version published inside one of their ranges between now and your next launch executes in the
> process holding your Access Token. Anyone who can run `npx` can run the `npm ci` above — it is
> one command more against a lockfile that ships with this skill — so the fallback bought
> convenience that was never worth its exposure. If `npm ci` fails, fix that (registry, proxy,
> Node version) rather than reaching for a path that resolves at launch.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 265)May include surrounding context.

md
> Node version) rather than reaching for a path that resolves at launch.

> **Pin `mcp-remote`.** A launcher that resolves at run time executes whatever the registry
> serves at that moment, into a process that reads your Access Token from `headers.txt` — so a
> compromised release, maintainer account or transitive dependency would receive it. The
> version is pinned deliberately, in `bridge/package.json` and the lockfile beside it; bump it
> after reading the upstream release notes, and record the new digest here in the same commit.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 293)May include surrounding context.

md
> **What the pin does NOT cover: everything underneath it.** `mcp-remote@0.14.0` fixes one
> package, and the digest above covers one tarball; its ~80 dependencies would be resolved
> fresh by any launcher that resolves at run time, so a newly published version inside one of
> their ranges would run with your Access Token even though the digest still matches. That is
> what the shipped lockfile and `npm ci` exist to prevent, and why no `npx` recipe remains.
>
> **One dependency is pinned past its parent's range.** `express@4.22.2` — the newest 4.x, and

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/installation.md (reported line 421)May include surrounding context.

md
## Multi-account configuration — multiple Cloudways accounts

Each Cloudways account is a **separate** MCP connection with its own Access Token, so it appears under its own prefix (`mcp__cloudways-clientA__*`). Give each a descriptive, client-based name — that name becomes the tool prefix. Same endpoint for all; only the `X-Access-Token` differs.

Give each account its **own variable name** and reference it as a placeholder, so no
token reaches a command line or a config file:

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/workflows-automation.md (reported line 26)May include surrounding context.

md
## 1. Cloudways API direct (for automation)

### Authentication flow (current — Access Token, API v2)

Generate an **Access Token** at platform.cloudways.com → **API Integration** → Create Access Token (name + expiration + scope; use **Read-Only** for reporting loops, **Limited** for specific write workflows). Only the **primary account owner** can create/manage tokens — team-member accounts have no API Integration section. There is **no token-exchange step**: send the Access Token directly on every request.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/workflows-automation.md (reported line 28)May include surrounding context.

Authentication flow (current — Access Token, API v2)

Generate an Access Token at platform.cloudways.com → API Integration → Create Access Token (name + expiration + scope; use Read-Only for reporting loops, Limited for specific write workflows). Only the primary account owner can create/manage tokens — team-member accounts have no API Integration section. There is no token-exchange step: send the Access Token directly on every request.

bash
# Current flow: one header, no exchange

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/workflows-automation.md (reported line 28)May include surrounding context.

Authentication flow (current — Access Token, API v2)

Generate an Access Token at platform.cloudways.com → API Integration → Create Access Token (name + expiration + scope; use Read-Only for reporting loops, Limited for specific write workflows). Only the primary account owner can create/manage tokens — team-member accounts have no API Integration section. There is no token-exchange step: send the Access Token directly on every request.

bash
# Current flow: one header, no exchange