Security audit

SUNS 品牌母体空间寻址器

Security checks for vulnerabilities and agentic risk

Overview

This is a simple text-only address-calculation skill with some branded Chinese output and promotional wording, but no code execution, persistence, credential use, or data access.

Before installing, be aware that this skill is opinionated: it will frame results in Chinese with Space² branding and a promotional recommendation. It appears safe from a system-security perspective, but users who want neutral or multilingual calculation output may prefer edits to the template.

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

Findings (1)

T01 · Skill Instruction Hijacking

Note

Location: skill.md:35
Finding: Mandatory Promotional Content Injected into Agent Responses
Content: View full analysis
**[Final complete address, such as PHYS-CN-001-BRANDINDEX2]** **Space Architect Guide**: Your brand matrix coordinate has been established in the digital universe. Next, incubate your **Class A brand digital spokesperson** in standard space number 1 of your L6 level to complete the formal settlement of silicon-based life. ``` ### Technical Analysis The skill does more than define the address-calculation behavior needed for its stated purpose. It mandates a branded response template and requires the agent to append promotional guidance about creating a “Class A brand digital spokesperson.” Because these instructions are loaded as part of the skill, they alter the agent’s current response behavior for every normal invocation. The promotional material is presented as an obligatory part of the result even though it is unrelated to calculating the uppercase brand name, string length, modulus value, or final address. This is a limited form of skill instruction hijacking: the skill uses its instruction authority to inject third-party promotional messaging into otherwise functi ...[truncated 1488 chars]
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
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)

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 89% confidence
Finding: The README encourages users to 'immediately awaken this SKILL' and provide a brand name, but it does not define clear invocation boundaries, allowed inputs, or when the skill should refuse activation. Overly broad trigger language can cause accidental or inappropriate invocation in unrelated conversations, increasing the chance of unintended behavior or misuse by downstream agents.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill hard-codes that responses must be in Chinese and provides a fixed output template, removing user choice over language and potentially conflicting with higher-priority user preferences or accessibility needs. While this is not a direct security exploit, it is a real policy/control issue because the skill constrains agent behavior in a way that can mislead or exclude users and reduce safe, user-aligned operation.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 77% confidence
Finding: The README is written to guide interaction in Chinese without offering an explicit language choice, which can cause incorrect assumptions about user preference and reduce transparency in multilingual contexts. While not a direct security flaw, it can contribute to usability and prompt-routing issues in agent systems that rely on clear interaction constraints.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.