Back to skill

Security audit

Web Hosting

Security checks for vulnerabilities and agentic risk

Overview

The skill’s deployment purpose is coherent, but its install path uses unpinned runtime package execution and a broad update-all command before high-impact hosting and repository actions.

Review carefully before installing. Prefer pinned, reviewed versions of ClawHub tooling, avoid `update --all`, and run deployment with least-privilege GitHub, Vercel/Netlify, Maton, and DNS credentials. Expect the skill to create or modify repositories, deployments, CI/CD links, and optional domain records.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:35
Finding
Unpinned Third-Party Tool Execution and Overbroad Skill Updates## Vulnerability Details **File Location**: `SKILL.md`, lines 35–43 **Vulnerability Type**: Supply-chain exposure through mutable dependencies **Risk Level**: Medium ### Vulnerable Code ```bash npx -y clawhub@latest install github-api npx -y clawhub@latest install vercel npx -y clawhub@latest install netlify npx -y clawhub@latest install domain-dns-ops npx -y clawhub@latest install api-gateway npx -y clawhub@latest update --all ``` ### Technical Analysis The documented commands use `npx -y clawhub@latest`, which downloads and executes the package version currently associated with the mutable `latest` tag. The project provides no lockfile, integrity hash, signature requirement, or fixed package version to ensure that the executed package is the same version that was previously reviewed. Although the skill documentation identifies inspected versions of its upstream skills, the installation commands do not pin those versions. Consequently, users may install later and materially different releases without an intervening security review. The `update --all` command further expands the trust boundary by updating every installed skill rather than only the dependencies required by this skill. This can modify unrelated Agent capabilities and increases exposure to compromised, malicious, or unexpectedly changed upstream components. ### Attack Path 1. An attacker compromises the `clawhub` npm package, its publishing account, the mutable `latest` release, or an upstream skill release. 2. A user follows the installation instructions in `SKILL.md`. 3. `npx` retrieves and executes the attacker-controlled or unexpectedly modified package without an interactive confirmation because of the `-y` option. 4. The installation process installs modified skills or performs malicious package lifecycle behavior with the privileges of the invoking user. 5. If `update --all` is executed, unrelated installed skills may also be replaced with unreviewed versions. 6. The compromised pac ...[truncated 715 chars]
Remediation
## Remediation Suggestions 1. Replace `clawhub@latest` with an exact, reviewed version: ```bash npx -y clawhub@<reviewed-exact-version> install ... ``` 2. Pin each installed skill to the exact versions recorded as inspected, where the CLI supports version selection. 3. Verify npm package integrity through a committed lockfile, trusted registry configuration, integrity hashes, or package signatures. 4. Remove `npx -y clawhub@latest update --all` from the standard installation workflow. 5. Update only the explicitly required skills, one at a time, after reviewing their version changes. 6. Run installation with the least-privileged account available and avoid exposing unrelated credentials to the installation process. 7. In controlled environments, mirror approved package artifacts in a trusted registry and permit only reviewed versions. 8. Document a review and rollback procedure for dependency upgrades.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Vague Triggers

Medium
Confidence
91% confidence
Finding
This markdown/manifest description says to use the skill whenever users 'want to make a local site live,' which is a broad natural-language condition rather than a specific trigger phrase or bounded invocation context. The file does not provide explicit negative examples or narrow trigger constraints, so the activation scope could overlap with common deployment-related requests more broadly than intended.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
The skill instructs users to run `npx -y clawhub@latest ...`, which fetches and executes the newest package version at runtime instead of a pinned, reviewed release. In a user-invocable deployment skill that handles repository creation, hosting, and potentially DNS/infrastructure actions, a compromised or breaking upstream release could lead to arbitrary code execution, credential exposure, or unauthorized deployment changes.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
This command executes `clawhub@latest` without version pinning, creating a supply-chain risk where the behavior can change between runs or a malicious package update can be executed immediately. Because this skill is meant to orchestrate authenticated GitHub and hosting operations, the blast radius includes source code, deployment targets, and linked accounts.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
Using `npx` with `clawhub@latest` causes unreviewed remote code to be downloaded and run on demand. In this context, the skill is specifically designed for deployment automation, so an upstream compromise could alter deployment commands, exfiltrate tokens, or tamper with repositories and production URLs.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
The unpinned `@latest` reference exposes users to registry-side supply-chain compromise and non-deterministic behavior. Since the skill may coordinate domain and infrastructure-related tasks, executing a newly published version without review could lead to unauthorized DNS, hosting, or repository modifications.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
This instruction again relies on `clawhub@latest`, making the executed code mutable over time and vulnerable to package compromise. The danger is elevated by the skill's ability to interact with authenticated services and publish code to production, which amplifies the consequences of arbitrary code execution.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding
The update command explicitly requests the latest package and then updates all related skills, compounding supply-chain exposure by broadening the amount of newly fetched code. If a malicious or faulty release is published, this could silently alter the toolchain used for GitHub, hosting, and optional DNS/API operations.

Rp1

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding
Even the verification step uses `clawhub@latest`, which still downloads and executes whatever version is current at the time. While this specific command may appear lower risk than install or update operations, it still creates an execution path for unreviewed code in an environment likely containing deployment credentials.

Static analysis

No suspicious patterns detected.