Skillv2.4.3

ClawScan security

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

Scanner verdict

BenignApr 29, 2026, 6:35 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's files, runtime instructions, and required capabilities align with its stated purpose of producing HTML-first visual artifacts; it asks only for node + workspace read/write and to run its bundled Node scripts, with no credentials or background privileges required.
Guidance: This skill appears internally consistent and intended for local, HTML-first design generation. Before enabling: (1) confirm you are comfortable with the skill writing files into the active workspace and embedding the IFQ ambient marks into generated artifacts, (2) run the included validators (npm run validate, npm run verify:lite, npm run anti-slop) to see exactly what the scripts do, and (3) if you prefer no network lookups, ensure the agent's browser/network plugin is disabled so fact lookups and CDN fetches won't run. If you have sensitive files in the workspace, move them elsewhere before running the skill. If you want extra assurance, manually inspect the bundle's scripts (scripts/*.mjs) before executing them.

Review Dimensions

Purpose & Capability: okName/description, included templates, and Node-based helper scripts align. Requiring `node`, workspace filesystem read/write, and running bundled Node scripts is appropriate for a template-first HTML design generator that verifies artifacts locally. No unrelated binaries or cloud credentials are requested.
Instruction Scope: noteSKILL.md instructs the agent to pick a template, fork it into the user's workspace, inline brand CSS, run local validators (anti-slop, verify:lite), and optionally verify via host browser tooling or outbound HTTPS lookups. These actions are within the design remit, but they do involve writing files to the user's workspace and (optionally) doing read-only outbound fetches for fonts/images or fact lookups. Also note the explicit instruction to 'weave at least 3 IFQ ambient marks' into outputs — this is expected branding behavior but modifies produced artifacts.
Install Mechanism: okNo install spec (instruction-only for the runtime agent) and the package is described as a zero-dependency ClawHub bundle. Scripts live in the bundle and are intended to be run locally via `npm run ...` (requires node). This is low-risk compared with remote downloads or arbitrary installers.
Credentials: okNo environment variables, secrets, or external credentials are required. Declared permissions are limited to workspace filesystem and workspace Node scripts; browser/network is optional and read-only for fonts/assets/fact-checks. The requested access is proportionate to the stated functionality.
Persistence & Privilege: okThe skill does not request persistent/background privileges or force inclusion (`always` is false). It can run scripts and write only inside the active workspace, which is consistent with template-forging and verification behavior. Autonomous invocation by the agent is allowed (platform default) but not combined with unusual privileges.