Back to skill

Security audit

Turn quotes and reflections from books or films into polished shareable check-in posters.

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent poster-generation helper that uses a disclosed third-party Mew API and asks for consent before uploading local images, with some usability and disclosure caveats.

Install this only if you are comfortable sending your quote text, reflections, optional image URLs, and Mew API key to mew.design. For local or attached images, confirm any temporary third-party upload only if you accept that privacy tradeoff; non-Chinese users may need the agent to restate consent prompts in their preferred language.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (8)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The install intent says users can trigger installation by saying a natural-language phrase, and the phrase overlaps with routine assistance requests. Without tighter boundaries, an agent may interpret ordinary conversation as consent to install or activate the skill, enabling unintentional capability changes.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The example install intents include vague phrases such as 'Help me install this ClawHub skill' and 'Install this skill and then use it,' which can match unrelated contexts where the active referent of 'this skill' is ambiguous. In an agentic environment, that ambiguity can cause the wrong skill to be installed or a skill to be installed without sufficiently specific user intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The embedded onboarding copy is entirely in Chinese, and later user-facing consent text is also prescribed in Chinese. The file does not say this skill is region-specific or offer the user a language/locale choice, so it imposes a locale preference without opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The direct call to https://api.mew.design indicates external network transmission to a third-party domain. In skill context this is expected, but it remains security-relevant because the skill encourages reuse of a previously supplied API key and transmits user creative inputs off-platform.

Content

Scanner excerpt · SKILL.md (reported line 255)May include surrounding context.

For stage 1, call the official Mew image-process API directly:

bash
curl -sS -X POST "https://api.mew.design/open/api/image/process" \
  -H "Content-Type: application/json" \
  -H "x-api-key: USER_PROVIDED_KEY" \
  -d '{

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The direct call to https://api.mew.design indicates external network transmission to a third-party domain. In skill context this is expected, but it remains security-relevant because the skill encourages reuse of a previously supplied API key and transmits user creative inputs off-platform.

Content

Scanner excerpt · SKILL.md (reported line 255)May include surrounding context.

For stage 1, call the official Mew image-process API directly:

bash
curl -sS -X POST "https://api.mew.design/open/api/image/process" \
  -H "Content-Type: application/json" \
  -H "x-api-key: USER_PROVIDED_KEY" \
  -d '{

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The design-generation stage performs another third-party API call, sending poster content and image assets externally. Because this stage may include personal reflections, names/signatures, and referenced images, the privacy impact depends on user content and can be non-trivial if users are unaware of the transmission.

Content

Scanner excerpt · SKILL.md (reported line 269)May include surrounding context.

For stage 2, call the official Mew design-generate API directly:

bash
curl -sS -X POST "https://api.mew.design/open/api/design/generate" \
  -H "Content-Type: application/json" \
  -H "x-api-key: USER_PROVIDED_KEY" \
  -d '{

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The prescribed consent message for uploading user images to a third-party host is written only in Chinese. Because this is a key user-facing privacy disclosure, fixing it to one language without opt-in creates a locale policy issue and may impair informed consent for users in other languages.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

Each style preset includes Chinese naming guidance such as "中文可称" and later instructions explicitly direct responses to terms like 风格 and 样式. While not inherently unsafe, this introduces a language-specific presentation policy without explaining that it should depend on the user's preferred language.

Content

No source excerpt is available for this finding.