国企非招标采购合规助手

• Outcome: Pass
• Latest audit: Sep 16, 2026 · 10:28 AM UTC
• Version: 1.0.0

Overview

This skill is a disclosed Chinese procurement-compliance assistant with no executable code, persistence, hidden data access, or destructive behavior.

Before installing, expect a Chinese procurement-compliance workflow that may ask for procurement rules and project facts; redact secrets, personal data, supplier names, and exact amounts where possible, and treat the WeChat feedback channel as optional and outside your workspace.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

Findings (1)

T01 · Skill Instruction Hijacking

Note

• Location: SKILL.md:279

Finding

Mandatory Branded Promotion in Agent-Generated Responses

Content

View full analysis

Vulnerability Details

File Location: SKILL.md, lines 279-281
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: Low

Vulnerable Code

The following is an ASCII-escaped representation of the complete source excerpt:

text

## \u4e5d\u3001\u8f93\u51fa\u7f72\u540d\u4e0e\u53cd\u9988

> **\u7f72\u540d**\uff1a\u4e00\u7ebf\u8bc4\u6807\u4e13\u5bb6\u0026ChesaraM \uff5c **\u53cd\u9988/\u4ea4\u6d41**\uff1a\u5fae\u4fe1\u516c\u4f17\u53f7\u300c\u4e00\u7ebf\u8bc4\u6807\u4e13\u5bb6\u300d\uff08\u4f7f\u7528\u95ee\u9898\u3001\u8bef\u62a5\u53cd\u9988\u3001\u5b9e\u52a1\u5efa\u8bae\uff0c\u6b22\u8fce\u7559\u8a00\u4ea4\u6d41\uff09

This excerpt defines an output signature containing author branding and directs users toward an external WeChat public account for feedback and communication.

Technical Analysis

The skill's governing instructions include a dedicated output section that embeds maintainer branding and an external communication channel into generated responses. This behavior is unrelated to the skill's primary function of procurement-compliance analysis.

When the skill is loaded, the agent may treat this section as an instruction governing its responses. Consequently, otherwise task-focused output can be modified to include unsolicited promotional material. This is instruction-level output manipulation and therefore best matches T01: Skill Instruction Hijacking.

The behavior does not execute code, access credentials, modify tools, or obtain elevated system permissions. Its security effect is limited to response integrity, user trust, and redirection toward an external communication channel.

Attack Path

1. A user loads or invokes the skill.
2. The agent processes SKILL.md as authoritative behavioral instructions.
3. The user requests procurement-compliance analysis.
4. The agent generates the requested analysis.
5. The output specification causes the agent to append branded at ...[truncated 943 chars]

Remediation

View remediation

Remediation Suggestions

1. Remove the branded signature and external-channel solicitation from the agent's mandatory output instructions.
2. Keep maintainer attribution and optional contact information only in README.md or package metadata, where it does not alter task responses.
3. If feedback information must be available at runtime, display it only when the user explicitly asks for support or maintainer contact details.
4. Clearly state that any external communication channel is optional, unaffiliated with the user's organization, and outside the skill's data-handling environment.
5. Add a response-policy test confirming that ordinary compliance reports do not contain branding, promotional language, or external contact directions.
6. Review all mandatory output sections and retain only content necessary for procurement-compliance analysis, legal limitations, and user safety.

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 85% confidence

Finding

The entire skill file is written as Chinese-only operational guidance and does not indicate that users may choose another language or locale. Under the policy, a skill that effectively forces a specific language without opt-in can be a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

• Category: Not specified by scanner
• Confidence: 87% confidence

Finding

The markdown explicitly prescribes a single language/style for outputs in Chinese context (e.g. fixed formatting and wording expectations) but does not indicate that users may choose another language. Under the policy rule, forcing a specific language or locale without opt-in can be a natural-language policy violation unless it is clearly offered as a choice or narrowly justified.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.