Skillv1.0.0

ClawScan security

Superdesign · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousMar 2, 2026, 4:33 AM

Verdict: suspicious
Confidence: medium
Model: gpt-5-mini
Summary: The skill's stated purpose (frontend design) is plausible, but its runtime instructions ask the agent to read large parts of the repo, fetch and obey external instruction files, and install/run an external CLI — behaviors that could exfiltrate code or be changed remotely and are not fully justified or declared.
Guidance: Before installing or running this skill, understand that it instructs the agent to: (1) scan and copy large parts of your codebase into .superdesign/init/, (2) install and run an external npm CLI (@superdesign/cli) and require you to 'login', and (3) fetch and obey additional instructions from raw.githubusercontent.com at runtime. These behaviors can upload sensitive source code and allow the remote documents or the CLI service to control what the agent does. Recommended precautions: (a) review the remote files (the two raw.githubusercontent URLs) yourself before allowing the skill to fetch/obey them, (b) inspect the @superdesign/cli package (npmjs/GitHub repo, maintainer, install scripts) in a safe environment, (c) do not run this on sensitive or private repositories without explicit approval and a legal/privacy review, (d) consider running the skill in an isolated container or sandbox, and (e) require explicit user confirmation before the agent performs repo-wide scans, installs, or uploads. If you need a less risky workflow, prefer tools that let you selectively send small, explicitly-chosen UI files rather than automatic full-repo extraction.

Review Dimensions

Purpose & Capability: noteThe skill is a design agent and legitimately needs UI code/context to produce pixel-accurate drafts. However, it mandates collecting FULL source code for many components/layouts and writing full init files into .superdesign/init/, and it also instructs installing a global npm CLI (@superdesign/cli) that is not described in the registry metadata. Fetching remote runtime instruction files (INIT.md, SUPERDESIGN.md) from an external raw.githubusercontent URL extends the skill's behavior beyond the local description.
Instruction Scope: concernSKILL.md requires reading the entire UI dependency tree, copying full source files (components.md, layouts.md, theme.md, etc.), creating .superdesign/init/ files, and then passing many local files to the external CLI. It also mandates fetching and following external runtime instructions (two raw.githubusercontent URLs) — giving the remote documents the power to change runtime behavior. These steps reasonably allow the CLI/service to receive large amounts of repository code (potentially sensitive) and permit remote instructions to alter agent actions.
Install Mechanism: concernThere is no declared install spec in the registry, but SKILL.md instructs the agent to run `npm install -g @superdesign/cli@latest` if the CLI is missing. Installing an unsigned/undeclared global npm package at runtime is a moderate-to-high risk vector (unknown package provenance, post-install scripts, etc.). The skill does not document verifying the package source or provide alternative vetted install methods.
Credentials: noteThe skill declares no required environment variables or credentials, but it requires the user/agent to run `superdesign login` and to ensure the CLI is logged in before use. That login step will create credentials/tokens (not declared or scoped here) and is necessary for the CLI to communicate with its backend. The lack of declared envs/tokens is not strictly inconsistent, but it means authentication will be interactive and the user should understand what account/service will receive uploaded context.
Persistence & Privilege: okThe skill does not request always-on or system-wide privileges and is user-invocable only. Its operations involve creating/reading files under the project (.superdesign/init/) which is expected for a repo-scanning design tool. It does not attempt to modify other skills or global agent settings.