Back to skill

Security audit

风格卡 · Style Card

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because its screenshot helper can silently install an unpinned package globally, while the scanners also report a mismatch between the declared purpose and some functionality.

Consider avoiding installation unless you are comfortable with the possibility of a silent global Playwright install and unpinned package use. Also consider whether external font requests and sending reference images or prompts to model providers are acceptable for your privacy needs. The scan covered all listed components, but static checks could not inspect binary image contents, so evidence about those files is incomplete.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied script’s primary function is browser-based screenshot export of an existing HTML document. It does not analyze reference images, construct or store structured style cards, select styles from a library, or generate styled content from text. Screenshot rendering could support a larger image-generation workflow, but the supplied chunk contains no connection to that declared workflow and explicitly targets HTML knowledge-card export. The dependency checks and installation are implementation details, not separate grounds for the mismatch.

Content

No source excerpt is available for this finding.

YARA rule 'privilege_escalation_tools': Privilege escalation tools and techniques [hacktools]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · examples/电影叙事排版风_demo_敢不敢裸辞.png (reported line 4006)May include surrounding context.

text
uRދ|L�aVL�V�;�
��&9�����3A�ހ�̪j��������?X{��]Μ��;����0.
1��R���>� Q���d�JC^#�wX���q���z<9�X��wc[Y�ɛH9�ll8Y�wγ3>w�E�|��[w�;>>[�,���~z`����tzzrr<\8.lk��ԟ`?G����ݕ��?���^��������U����l��c�x�����{�F�-Z4�w���%l����ݽ���.��z����{�9#���ۋ�T��F�������'����}��sѽ�:�ծ�����-i�ŝڌU�cYt���mH���DG�T�
g��
��U)ײ&&g6)K�Ho����l�W�����-߼uS��$���?pspy��!�_�6mvh�^4Pg��l�\������i�梹%�H���(���Μ��������;�[��<�=����P����bm?v��rg���B��2dx�r�\u�&]c����}�ͥ�/�B�Xps�X�����⼶~���i�R�@�d�UQ�:�s{~�F2���O��s-udOg$Ī"Q�9���ٸ\�:�5�cꛣ����_C9�CoS��l��u����R;h��/�č����L���8�AY�F���p�D��$�(5��tb����F�y������ԗVǧ[��N���c�Q�ا������l̕��,�����r�,-W{�ץ��hTcDN�Q�u�2˅���B������4@5�좖k����Ic�@���I\�xBC��k���p�$s**�WFw�ݺ�SQ ٯ�ȼ�
& �n�*Tߪe����d��X�n���+L�����]����Ws���i�R��΀7;�ܳ���繼�5�����p$ݮ�F��g�,Z�%˔X���B	�7&�l�i�揩E}�	(�TWx:~��[�4Qҧ�q�$U�%og�ˌ����J_�lKV�qj}�����

The HTML workflow requires Google Fonts without warning users about third-party network requests.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

Requiring remotely loaded Google Fonts introduces third-party requests during rendering without disclosing that network dependency. These requests can expose the renderer's IP address and request metadata, although the supplied content does not establish transmission of image contents or user text.

Content

No source excerpt is available for this finding.

The typography guidance directs window titles and brand lines to use English without user opt-in.

Medium
Category
Output Handling
Confidence
89% confidence
Finding

L012 directs the agent to use English for window titles and branding because it looks better and produces fewer spelling errors. This imposes a language choice on part of the generated output without asking the user or making the guidance conditional on their preferred language.

Content

No source excerpt is available for this finding.

Reference images are sent to a model provider without a privacy warning.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow instructs users to send the reference image and prompt to services such as Claude, GPT-4V, or Gemini, but does not warn that images may contain confidential or personal information disclosed to that provider. The later claim that no external API is needed does not explain or remove this external processing.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx playwright'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

The dependency check can silently install Playwright globally.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The dependency check can perform a global package installation while redirecting both output streams to /dev/null. This hides a consequential host modification inside a screenshot operation and prevents users from readily observing installation activity or warnings.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx playwright'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

Although Playwright is a CLI rather than an MCP server here, the dependency check invokes an unpinned package and falls back to an unpinned global installation. When package resolution fetches a compromised release, package or installation code can run with the invoking user's privileges.

Content

No source excerpt is available for this finding.

The screenshot helper can install Playwright globally, modifying the host's shared Node.js environment.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

A failed version check triggers npm i -g playwright, allowing an image-export operation to modify the user's shared Node.js installation. This exceeds the necessary scope of rendering and can introduce executable dependencies or disrupt other workflows without separate approval.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx playwright'.

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The screenshot command executes Playwright without constraining its version or requiring a reviewed local installation. Depending on the environment, npx can resolve or fetch an unreviewed package version, creating a supply-chain execution risk rather than an MCP-specific issue.

Content

No source excerpt is available for this finding.

L49 and L58 require English metadata text without user opt-in.

Low
Category
Output Handling
Confidence
98% confidence
Finding

The typography instructions specify uppercase English for the metadata row, and the content slot repeats the English requirement. This imposes a natural-language constraint on generated content without offering a language choice or explaining why English is necessary.

Content

No source excerpt is available for this finding.

L87 requires a Chinese main title without offering a language choice.

Low
Category
Output Handling
Confidence
97% confidence
Finding

The reusable generation prompt explicitly calls for a Chinese pixel-font title. The RPG visual style does not inherently require Chinese, and the file provides neither user opt-in nor a justification for restricting the title's language.

Content

No source excerpt is available for this finding.

L49 and L58 mandate English for the metadata row without user opt-in.

Low
Category
Output Handling
Confidence
98% confidence
Finding

The style card requires uppercase English metadata and reinforces that requirement in its content slots. It does not offer a language choice or explain why the Y2K aesthetic requires English text.

Content

No source excerpt is available for this finding.

L57 restricts the top metadata slot to English without user opt-in.

Low
Category
Output Handling
Confidence
97% confidence
Finding

The top metadata slot explicitly specifies an English command-style phrase. This constrains generated text to a particular language without providing a user choice or a documented language-specific justification.

Content

No source excerpt is available for this finding.

The helper claims to install Chromium automatically but never performs browser installation.

Low
Category
Excessive Agency
Confidence
98% confidence
Finding

L04 promises automatic first-run Chromium installation, and L24 says the following code ensures Chromium is available. However, L25 only checks or installs the Playwright CLI, and L26–L29 invoke screenshot capture without running a browser-install command; a clean environment can therefore fail despite the documented setup guarantee.

Content

No source excerpt is available for this finding.