ClawHub
Back to skill

Security audit

Pet Food Compare

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only pet-food comparison skill with disclosed public-source research behavior and no hidden code, persistence, or credential access.

Installers should expect this skill to perform public web, image, and review-source research about pet foods and to output in Simplified Chinese. Treat the animal-condition mapping as informational product context only, especially for prescription or health-related diets; final pet health decisions should rely on a veterinarian.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The manifest frames the skill as one for comparing multiple pet food options when a user is choosing among them. But the file explicitly requires a full single-product report and says not to omit fields such as risk additives, product incidents, comment frequency, animal-condition matching, and packaging images even when only one product is provided. That behavior is broader than comparison and turns the skill into a general product dossier generator.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest says the skill only performs neutral information collection, summarization, and comparison, and must not make recommendation decisions. However, this section instructs the model to map product features to '对应动物情况' such as high activity, obesity, allergy elimination, skin/coat concerns, and urinary issues, which goes beyond raw comparison into suitability-oriented guidance. Even though the text says this is 'not recommendation,' the implemented behavior still provides decision-support judgments about which animals a food corresponds to.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The checklist explicitly requires that the entire output be in Simplified Chinese. This is a natural-language locale restriction that applies regardless of user preference, and the file does not offer an opt-in or alternative language path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The output specification states that the skill's output language must be Simplified Chinese because its main users are mainland China users. This forces a specific locale without documenting user opt-in or a justified hard regional constraint.

Content

No source excerpt is available for this finding.