Security audit

巴别渊

Security checks for vulnerabilities and agentic risk

Overview

This is a creative text-transformation skill with no code, tools, persistence, or data access beyond the text the user asks it to rewrite.

Install this if you want an intentionally destructive, surreal rewriting mode. Be aware it is designed not to preserve meaning unless you explicitly ask for recognizability, so it is unsuitable for faithful editing, summarization, or technical work.

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
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (3)

Anti-Refusal Statement

High

Category: Anti-Refusal
Content: - favor semantic drift over fidelity - allow first-person presence to appear as a cracked observing voice - occasionally introduce synesthetic crossings, sensory miswiring, and conceptual resonance - let imagery mutate across domains without warning - preserve only a faint residue of the original input unless the user explicitly asks for more recognizability - keep the result vivid, dense, and textural rather than merely random - avoid neat aphorisms, joke formatting, meme cadence, or tidy surrealism
Confidence: 80% 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.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: The activation condition is intentionally broad and subjective, using phrases like 'clearly wants' various degradation styles rather than limiting invocation to explicit triggers. This can cause the skill to activate when a user did not clearly request semantic corruption, leading to unintended override of normal assistant behavior and degraded output quality.

Natural-Language Policy Violations

Medium

Confidence: 90% confidence
Finding: The interface metadata forces a Chinese display name and description without indicating language flexibility or user choice. In practice this can create usability and routing issues for users in other locales, and may cause accidental invocation mismatches or confusion about the skill's purpose.

Static analysis

No suspicious patterns detected.