Back to skill

Security audit

Aesthete Agent — Frontend Aesthetics, Proof-Tested

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious with medium confidence, chiefly because scanners report capabilities and installation instructions that do not consistently match its stated offline color-tool purpose.

Consider the skill’s unpinned installation examples, reported purpose mismatch, and incomplete scan coverage before installing. The report does not establish malicious intent or a current compromise, but it also does not provide complete evidence to treat the skill as safe.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (101)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied module's primary purpose is typography: its build and emit functions calculate type scales and produce typography specimens and tokens. It performs no palette generation, contrast calculation, or color-accessibility evaluation; the few fixed colors merely style the specimen. Although typography broadly relates to frontend aesthetics, the description explicitly excludes pure text layout, making this a material scope mismatch rather than a supporting color implementation detail. The code uses only standard-library functionality and local file writes, with no observed network access or external dependency.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Although title typography is adjacent to frontend aesthetics, this chunk's primary function is font subsetting, a distinct capability not disclosed in the detailed color-focused description. It performs no palette generation, color analysis, or accessibility checks. Its required fonttools and brotli imports also directly contradict the stated pure-stdlib implementation. The code remains offline: the external URL is only printed as manual download guidance.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The primary purpose matches, and the visible resource accesses—local palette libraries, profiles, images, HTML/SVG files, and related scripts—are consistent with that purpose. No network access or unrelated activity is evident. Nevertheless, the dependency, preference-learning, and perceptual-contrast claims materially overstate what the supplied implementation provides. The code chunk ends during argument-parser setup, so capabilities delegated to unprovided modules and the exact library size cannot be verified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The tested functionality concerns text layout rather than color generation or validation. Although typography can support frontend aesthetics, this chunk independently exercises a typography engine and typography-only audits, which conflicts with the description's explicit exclusion of pure text layout. The palette passed to page_audit does not change the typography-specific assertions. The visible code is a test suite rather than the underlying implementation; its local temporary-file operations and documentation reads are ordinary test support, not separate security concerns.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The only supplied code is an installation command, not an implementation of the declared color tools. It invokes an external CLI to add a remotely hosted skill, which is not disclosed in the description and conflicts with its offline positioning. Although the command appears in a README and could be setup documentation, no underlying skill implementation is supplied to verify the claimed palette capabilities.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

These tests target typography functionality rather than color generation or validation. Although typography broadly relates to frontend aesthetics, the description specifically frames the skill around color and does not disclose a font asset pipeline or font-weight auditing. The auto-subsetting test also explicitly checks for fontTools, contradicting the pure-stdlib claim for that path. The implementation modules are not supplied, so the evidence establishes the capabilities exercised by the tests rather than their internal implementation. Temporary-directory writes are ordinary test behavior, not independently suspicious.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

The showcase generation itself matches the declared frontend-color preview purpose, and its local file operations are supporting implementation details. The material mismatch is the representation of previews as mechanically validated: badge() hard-codes safety successes, wall_card() can assert AA without meeting its normal-text threshold, and the footer asserts comprehensive gate passage without demonstrated enforcement. Imported helpers may provide additional color processing, but the supplied code does not establish the unconditional validation claims it publishes. Other advertised features need not appear in this preview-focused module.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The supplied chunk performs plugin installation or registration rather than the declared color-processing functions. This changes a tool profile and introduces an external source, neither of which is disclosed in the description. Although installation can be legitimate setup, this chunk provides no evidence of the advertised offline implementation. The contents and subsequent behavior of the referenced plugin cannot be determined from this command.

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 146)May include surrounding context.

md
| 4 render evidence | `render_probe.js` + `lint_page --render` | measured theme/dosage/contrast/font |

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 · docs/img/showcase-mobile.png (reported line 266)May include surrounding context.

text
>]������ �tqz5�A;%%������z�R�h43߻4GZː�:"��*P#rg0&r=ց�\2��l̔&���!CVi>f6��>y+�s �r*�X�A�@���T��*�,��o�\�uw?u~lb���a"Z��'N�}p��s':‚lv�ԉ�t��WP,d/�������膭���/�yN��E/��0��`p0ӽbA�'/���G��s/;=y����Dz,��A��/
�
_D�� s����I���DWJ&7�\,"��"�X��@�Q��Ǚ�,�T���Cn�.	 OS�1U��8��	:Z�Ұ��:[��Q6t�r� K�F��&[�?�*�";��\cm��(���-� MU�/]��K����,Q��S:��.��PSpYFC��[<��xL�
�w�=�zazz�R__�5Fi05u���?������t�p&3r����kY��8j<�a���m����ї��[���Md
>�Ѧ�����ޙ ��,��]�^2#�>���zv���EkZ7?��p��Z:3x�����WN�ɹ1�8s�G_�ϗ�Q���_�𮮆(�@#�	S�yzJx�����{�������!-
���90���@էR��52i��oD?���l�+�A�r9��)ރ��E��*�@�*߂ޥ�FA#�L@� ȝ�$��z�%��ۡk>�R,����Z�l�tP���V�cD5�@e��Dz24	�U�G��U�0�YAU��R\�bM��%�y<au�m���7'��ɱS�������Ƀ����w����N��^=y~�{�'��ϝ��M{Q�f���5щ7������N��T6��č��_��S����#�W' �y��GH|��乗�����߽p������Gh������GN������
�g����~P�f�(�������ʛ����?��6.�Q�"�	�v.��d��*X��S(�̌���K���Yf�$Нb�

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · palettes/palettes.json (reported line 4)May include surrounding context.

json
{
 "_meta": {
  "schema": 2,
  "description": "Curated palettes harvested from public palette community charts (trending/popular/latest, 48 each; style-filter URLs proved to be SEO fallback clones of popular — do not use). tier = crowd signal (3>=50K likes, 2>=20K, 1>0, 0 new/unranked); occurrences = cross-page recurrence; computed_styles from HSL stats; flags from objective gates (CB-COLLISION-RISK / SLOP-SIGNATURE / NARROW-LIGHTNESS). Gates reject only broken architecture, never judge beauty — final judgment happens in context at use time.",
  "harvested": "2026-10-15",
  "counts": {
   "kept": 119,

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · scripts/release_check.py (reported line 40)May include surrounding context.

python
@check("C1 测试全绿")
    def c1():
        env = dict(os.environ, AESTHETE_RC_CHILD="1")
        r = subprocess.run([sys.executable, "-m", "unittest", "discover", "-s",
                            os.path.join(tree, "tests")],
                           capture_output=True, text=True, timeout=600, env=env)

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_v02300.py (reported line 21)May include surrounding context.

python
if os.environ.get("AESTHETE_RC_CHILD"):
            self.skipTest("release_check 子进程内,防递归套娃")
        r = subprocess.run([sys.executable, RC], capture_output=True, text=True,
                           timeout=600, cwd=ROOT, env=dict(os.environ, AESTHETE_RC_CHILD="1"))
        self.assertEqual(r.returncode, 0, r.stdout + r.stderr)
        self.assertIn("裁决: PASS", r.stdout)

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The quick-install command executes the skills CLI without specifying a reviewed version and installs skill content from a mutable repository reference. A compromised future release could execute code or introduce adversarial agent instructions; this is a supply-chain exposure, not evidence of an existing compromise.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The installation example leaves both the skills CLI version and repository revision unspecified. Although this CLI is not itself an MCP server, the command still creates a real supply-chain exposure if a later package release or repository update is compromised.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The command can fetch and execute an unspecified version of the ClawHub CLI and installs a skill without an explicit immutable version. A compromised upstream update could therefore gain installer privileges or alter subsequent agent behavior, despite the palette engine's intended offline operation.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'uvx aesthete-agent-mcp'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The uvx command launches the MCP package without selecting a reviewed exact version. If package resolution supplies a compromised release, that server can execute with the launching user's permissions; offline palette processing does not eliminate installation-time supply-chain risk.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @smithery/cli'.

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The alternative MCP installation command runs an unpinned Smithery CLI, with -y suppressing the npx installation confirmation. A compromised release could execute with the user's privileges or modify the client's MCP configuration; no compromised release is demonstrated here.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'uvx aesthete-agent-mcp'.

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

This explanatory paragraph endorses the same unpinned uvx launch command rather than executing it directly. If used as launch guidance, it leaves the server artifact mutable and exposes the client environment to a compromised future package release.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The platform table repeats an installation command that does not constrain the ClawHub CLI or skill release. Following it trusts mutable upstream artifacts, creating conditional code-execution and agent-instruction supply-chain risks.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'uvx aesthete-agent-mcp'.

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The PyPI table entry recommends launching the MCP server without an exact version constraint. This creates a conditional supply-chain vulnerability because a later compromised release could run arbitrary code with the server process's permissions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx -y @smithery/cli'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The abbreviated table example again recommends an unversioned Smithery CLI with automatic confirmation. Although the ellipsis makes this an incomplete command, it reinforces an installation approach that could execute an unreviewed future release.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The quick-start command runs the skills installer without a fixed package version and references a mutable skill repository. A compromised future release could execute with the installing user's permissions; this is a supply-chain exposure, not evidence that the current package is malicious.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation instructions invoke an unversioned npm CLI and a mutable repository reference. Following them can execute a different installer or install different skill content over time, exposing users to compromised upstream updates.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The command invokes clawhub without a fixed npm package version and installs a skill without an immutable release reference. A compromised upstream update could therefore change the code or agent instructions users receive.

Content

No source excerpt is available for this finding.