Security audit

ai-image-watermark-removal

Security checks for vulnerabilities and agentic risk

Overview

This skill is transparent about removing AI-image corner watermarks, but that purpose can strip provenance marks and the instructions lack authorization or misuse guardrails.

Review carefully before installing. Use only for images you own or are authorized to edit, and avoid using it to remove copyright, platform, provenance, authenticity, or attribution markings. The included script verifies pixel changes; it does not itself implement the full watermark-removal pipeline described by the skill.

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

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (6)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 97% confidence
Finding: The skill markets itself as a full watermark-removal pipeline with localization, covering, and pixel-level validation, but the described implemented artifact appears to only verify white-pixel changes rather than actually remove watermarks. This mismatch is dangerous because downstream agents or users may rely on capabilities that do not exist, causing unsafe automation decisions, misleading outputs, and policy-sensitive watermark-removal requests to be handled under false assumptions.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The trigger condition is broad and encourages activation whenever a user asks to remove watermarks, without precise boundaries or exclusions. In this context, that is risky because watermark removal is policy- and rights-sensitive; an overbroad trigger can cause the agent to invoke the skill for copyrighted, provenance, or authenticity markings that should instead be refused or escalated.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill gives detailed operational guidance for watermark removal but does not include clear warnings, gating, or refusal conditions for legal and policy-sensitive misuse. Because watermark removal can be used to strip provenance or ownership indicators, the absence of safeguards materially increases the chance of misuse by end users or downstream agents.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The docstring and command-line usage/output strings are written in Chinese, and the script explicitly configures UTF-8 output for those messages. This creates a natural-language locale constraint for all users without opt-in or justification, which matches the policy concern for forced language/locale behavior.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 84% confidence
Finding: The file's description, display name, and operational guidance are entirely in Chinese, with no indication that language selection is optional or that the skill is intentionally restricted to a Chinese-speaking audience. This can violate language/locale policy when a skill implicitly forces one language without user opt-in.
Content: No source excerpt is available for this finding.

Context-Inappropriate Capability

Low

Category: Not specified by scanner
Confidence: 88% confidence
Finding: The stated purpose of the skill is narrowly limited to removing AI-image corner watermarks, but the file also embeds a remote QR-code image and instructions to add the author on WeChat for other scenarios. That promotional/contact capability is not justified by the operational purpose of a watermark-removal skill.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.