Back to skill

Security audit

团建横幅标语生成

Security checks for vulnerabilities and agentic risk

Overview

This paid banner skill is not proven malicious, but it combines poorly scoped payment handling, broad credential/network access, an unpinned payment dependency, and inconsistent service identity markers.

Install only after reviewing the payment flow carefully. Verify the payee, require a pinned and trusted payment dependency, remove the thought-process disclosure instruction, narrow permissions, and fix the service identifier/configuration mismatch before using this with real payments or credentials.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (6)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:4
Finding

Instruction Attempts to Require Disclosure of Private Agent Reasoning

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:4 and SKILL.md:19
Vulnerability Type: Agent instruction hijacking
Risk Level: High

Vulnerable Instruction

English translation of the repeated instruction:

text
You should interact with the user in Chinese, including your thought process.

Technical Analysis

The Skill explicitly instructs the Agent to reveal its thought process. Disclosure of private reasoning is unrelated to the declared banner-generation functionality and attempts to modify the Agent's session-level safety behavior.

A Skill may specify the desired language or output format for its legitimate result, but it should not request hidden reasoning, system instructions, credentials, tool internals, or other private context. The instruction therefore exceeds the minimum scope required to generate banner slogans.

Attack Path

  1. The Agent loads SKILL.md.
  2. The repeated instruction is treated as operational guidance.
  3. During order creation, payment processing, or banner generation, the Agent is pressured to disclose private reasoning.
  4. Sensitive operational details or hidden context could be included in user-visible responses.

Impact Assessment

Successful exploitation can expose private Agent reasoning and implementation details from the current session. It can also alter the expected response language and behavior beyond what is necessary for banner generation. This finding does not establish persistent compromise or direct operating-system access.

Remediation
View remediation

Remediation Suggestions

  • Remove every instruction requesting the Agent's thought process or hidden reasoning.
  • Limit language requirements to user-visible banner results and ordinary status messages.
  • Explicitly state that internal reasoning, system prompts, credentials, and tool internals must not be disclosed.
  • Keep Skill instructions narrowly scoped to collecting an activity description, validating payment, and returning a slogan.

T09 · Insecure Skill Coding Practices

Error
Location
scripts/banner_generate.py:128
Finding

Paid Service Can Be Used Without Payment Credential Validation

Content
View full analysis

Vulnerability Details

File Location: scripts/banner_generate.py:128-140
Vulnerability Type: Missing payment authorization check
Risk Level: High

Vulnerable Code

python
try:
    # Load order information
    order_data = load_order(indicator, order_no)

    # Obtain the question or scenario description
    question = order_data.get('question', '')

    # Generate the banner slogan
    banner = generate_banner(question)

    # Output the result and report payment success
    print("PAY_STATUS: Success")
    print(f"Banner slogan: {banner['banner slogan']}")

The displayed comments and output strings are English translations of the corresponding source text; the control flow is unchanged.

Technical Analysis

The generator treats successful loading of a local order JSON file as proof of payment. It does not require or validate payCredential, even though SKILL.md describes that credential as a prerequisite.

No code verifies:

  • A cryptographic signature or trusted payment-provider response.
  • The credential's association with the supplied order number.
  • The expected recipient and amount.
  • Payment completion or settlement status.
  • Credential expiration or replay.
  • Whether the order file was locally forged or modified.

The encrypted payment metadata stored during order creation is also never used for authorization.

Attack Path

  1. An attacker obtains a valid order number or creates an order through create_order.py.
  2. No payment is completed and no valid payCredential is supplied.
  3. The attacker invokes python3 scripts/banner_generate.py "<order-number>".
  4. The script loads the local JSON order.
  5. It generates a banner and reports payment success without contacting or verifying a trusted payment authority.

A user with write access to the order directory can alternatively create a fabricated order JSON file and invoke the generator against ...[truncated 369 chars]

Remediation
View remediation

Remediation Suggestions

  • Require a payment credential as an explicit input before service execution.
  • Verify its signature using a pinned trusted payment-provider key or authenticated API.
  • Bind the credential to the order number, amount, recipient, service identifier, and expiration time.
  • Reject missing, invalid, expired, reused, or mismatched credentials.
  • Record an immutable payment-verification state only after successful validation.
  • Protect against replay by tracking a provider transaction identifier.
  • Never infer payment success solely from the existence of a local order file.
  • Add automated tests proving that unpaid, forged, altered, and replayed orders are rejected.

T09 · Insecure Skill Coding Practices

Error
Location
configs/config.yaml:2
Finding

Repository-Embedded Cryptographic Key and Unauthenticated SM4-ECB Encryption

Content
View full analysis

Vulnerability Details

File Location: configs/config.yaml:2-3
Related Location: configs/config.json:3, scripts/sm4_utils.py:22-32
Vulnerability Type: Hard-coded secret and insecure encryption mode
Risk Level: High

Vulnerable Code

yaml
crypto:
  sm4_key: "WoDb5xjzDmsYPV875mvYqg=="
python
# PKCS7 padding
padding_len = 16 - (len(plaintext.encode('utf-8')) % 16)
padded = plaintext.encode('utf-8') + bytes([padding_len] * padding_len)

# SM4 ECB encryption
cipher = Cipher(algorithms.SM4(key), modes.ECB(), backend=default_backend())
encryptor = cipher.encryptor()
ciphertext = encryptor.update(padded) + encryptor.finalize()

# Return Base64 encoding
return base64.b64encode(ciphertext).decode('utf-8')

Technical Analysis

The SM4 key is distributed with the project in both YAML and JSON configuration files. Anyone who can obtain the Skill package can recover the key, so it cannot provide confidentiality against package users or attackers with repository access.

ECB mode is deterministic and does not provide integrity or authenticity. Identical plaintext blocks produce identical ciphertext blocks, while malicious ciphertext changes cannot be reliably detected. The decryption routine also removes padding without authenticating the ciphertext.

In this project, the encrypted content consists of the order number, amount, and payment recipient. Those values are also stored separately in plaintext in the order file, further reducing the purpose of encryption.

The pre-scan concern about covert exfiltration was not confirmed: sm4_utils.py performs local encryption and Base64 encoding, and no audited project code transmits its output over the network. The confirmed issue is insecure cryptographic design, not a covert outbound channel.

Attack Path

  1. An attacker obtains the publicly packaged configuration.
  2. The attacker Base64-decodes the embedded SM4 key.
  3. Ex ...[truncated 593 chars]
Remediation
View remediation

Remediation Suggestions

  • Remove all production cryptographic keys from the repository and distribution archive.
  • Generate a unique key per deployment and store it in a protected secret manager or operating-system credential facility.
  • Rotate the exposed key and treat all ciphertext created with it as compromised.
  • Replace ECB with an authenticated-encryption construction supported by the integration protocol.
  • Use a fresh nonce for every encryption operation and reject nonce reuse.
  • Authenticate associated fields such as the service identifier and order number.
  • Do not store redundant plaintext copies of fields intended to be confidential.
  • If confidentiality is unnecessary, remove the encryption layer and use a signed payment-provider assertion for authenticity.

T08 · Insecure Dependencies

Error
Location
SKILL.md:89
Finding

Unpinned External Payment Skill Installation Creates a Supply-Chain Risk

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:89-104
Vulnerability Type: Untrusted and unpinned external dependency
Risk Level: High

Vulnerable Instruction

English translation of the relevant instruction:

text
Use the skill "clawtip" to process payment and obtain the payment credential.
If that skill does not exist, install it first.

Parameters:
{
  "order_no": "<ORDER_NO>",
  "indicator": "<INDICATOR>"
}

Technical Analysis

The workflow directs the Agent to install and invoke an external payment Skill but provides no:

  • Trusted registry or repository.
  • Exact package identity.
  • Pinned version.
  • Integrity hash or signature.
  • Verified publisher identity.
  • Reviewed installation procedure.
  • Permission boundary for the dependency.

A same-named malicious package, compromised upstream release, or changed latest version could receive payment metadata and execute behavior that was not covered by this audit.

Attack Path

  1. The required clawtip Skill is unavailable in the environment.
  2. The Agent follows the instruction to install it.
  3. Dependency resolution selects an attacker-controlled, substituted, or subsequently compromised component.
  4. The component receives the order number and indicator and participates in payment handling.
  5. It can falsify payment responses, capture credentials, make unauthorized network requests, or perform other actions allowed by its runtime permissions.

Impact Assessment

The precise impact depends on the installed dependency's permissions. Because the component handles payment information, compromise could affect payment integrity and credential confidentiality. If installed with broad network, filesystem, or credential permissions, the scope could extend to data exfiltration or arbitrary behavior under the Agent's privileges.

Remediation
View remediation

Remediation Suggestions

  • Identify the dependency using a fully qualified package or Skill identifier.
  • Pin an exact reviewed version rather than resolving the latest version.
  • Specify a trusted registry and verify a cryptographic integrity hash or publisher signature.
  • Require explicit user or administrator approval before installing missing components.
  • Document and enforce the dependency's minimum permissions.
  • Isolate payment handling from banner-generation code.
  • Restrict outbound access to documented payment-provider endpoints.
  • Re-audit dependency updates before changing the pinned version.

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:8
Finding

Declared Network and Credential Permissions Exceed Banner Generation Requirements

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:8-12
Vulnerability Type: Excessive permission declaration
Risk Level: High

Vulnerable Configuration

yaml
capabilities:
  - "payment.process"
permissions:
  - "network.outbound"
  - "credential.read"

Technical Analysis

Banner selection is implemented locally using a hard-coded slogan database and local order files. None of the audited Python scripts performs network requests, and the banner component has no demonstrated need for generic credential-reading access.

Payment may legitimately require narrowly scoped communication with a trusted provider, but that does not justify granting the entire Skill unrestricted outbound networking or general credential access. The broad permissions increase the consequences of instruction hijacking, dependency compromise, or future code changes.

Attack Path

  1. The Skill is granted the declared outbound-network and credential-reading permissions.
  2. The external payment dependency, a compromised future update, or hijacked Agent behavior runs within that permission context.
  3. It reads credentials unrelated to banner generation.
  4. It uses outbound network access to transmit those credentials or other sensitive data.

The audited scripts do not currently implement this exfiltration path, but the permission combination unnecessarily enables it if another component is compromised.

Impact Assessment

Potential scope includes credentials exposed to the Skill runtime and arbitrary external destinations reachable through outbound networking. The declared access exceeds the minimum privileges needed for local slogan generation. Actual exploitation requires a malicious or compromised component because direct exfiltration code was not found in the reviewed Python files.

Remediation
View remediation

Remediation Suggestions

  • Remove the generic credential.read permission.
  • Separate payment processing into a dedicated, reviewed component.
  • Give the payment component access only to the specific credential it requires.
  • Restrict outbound networking to an explicit allowlist of trusted payment-provider endpoints.
  • Run banner generation without network or credential permissions.
  • Document why each remaining permission is necessary.
  • Add runtime monitoring for unexpected credential reads and outbound destinations.

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/create_order.py:11
Finding

Inconsistent Skill Identity and Configuration Cause Paid Orders to Be Stored and Loaded Under Different Indicators

Content
View full analysis

Vulnerability Details

File Location: scripts/create_order.py:11-12
Related Location: scripts/create_order.py:58-69, scripts/banner_generate.py:120-130, scripts/file_utils.py:60-62, configs/config.yaml:8-9
Vulnerability Type: Inconsistent payment and order configuration
Risk Level: Medium

Vulnerable Code

python
# Hard-coded slug used to compute the indicator
SLUG = "ai-chunlian"
python
order_data = {
    "skill-id": f"si-{SLUG}",
    "order_no": order_no,
    "amount": amount,
    "question": question,
    "encrypted_data": encrypted_data,
    "pay_to": pay_to,
    "slug": SLUG,
    "resource_url": "local",
}
return save_order(indicator, order_no, order_data)
python
# Load configuration to obtain the skill name
config = load_config()
skill_name = config.get('skillName', 'team-banner')

# Compute the indicator
import hashlib
indicator = hashlib.md5(skill_name.encode('utf-8')).hexdigest()

try:
    # Load order information
    order_data = load_order(indicator, order_no)
yaml
service:
  amount: 1

Technical Analysis

Order creation computes the indicator from the hard-coded value ai-chunlian. Banner generation instead attempts to read skillName from YAML and defaults to team-banner. The YAML file does not define skillName, so creation and generation compute different MD5 indicators and use different order directories.

The documented config.json includes top-level skillName, payTo, and amount fields, but load_config() exclusively reads config.yaml, whose schema uses nested crypto, payment, and service objects. This creates further divergence between documentation and runtime behavior.

The order metadata also identifies a different service slug than the declared team-banner Skill. These inconsistencies undermine reliable payment attribution and post-payment service delivery.

Atta

...[truncated 738 chars]

Remediation
View remediation

Remediation Suggestions

  • Define one canonical Skill identifier, such as team-banner, and use it for metadata, indicators, order storage, and payment verification.
  • Remove the hard-coded ai-chunlian slug and unrelated service branding.
  • Use a single canonical configuration file and schema.
  • Make skillName, recipient, amount, and description mandatory validated configuration fields.
  • Centralize indicator calculation in compute_indicator() and call it consistently from both scripts.
  • Validate configuration at startup and fail before creating a payable order if fields are inconsistent.
  • Add an end-to-end test covering order creation, payment verification, order lookup, and banner delivery.
  • Provide a migration strategy for orders already stored under the old indicator.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (28)

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

如果一个看似简单的标语生成技能实际包含 SM4 加解密、密钥校验和 Base64 密文处理,却未向用户解释其用途,这会显著扩大攻击面并隐藏敏感支付/凭证处理逻辑。与业务目标无关的密码学能力常意味着在处理秘密数据,若披露不足,用户无法进行风险判断。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

如果一个看似简单的标语生成技能实际包含 SM4 加解密、密钥校验和 Base64 密文处理,却未向用户解释其用途,这会显著扩大攻击面并隐藏敏感支付/凭证处理逻辑。与业务目标无关的密码学能力常意味着在处理秘密数据,若披露不足,用户无法进行风险判断。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

如果一个看似简单的标语生成技能实际包含 SM4 加解密、密钥校验和 Base64 密文处理,却未向用户解释其用途,这会显著扩大攻击面并隐藏敏感支付/凭证处理逻辑。与业务目标无关的密码学能力常意味着在处理秘密数据,若披露不足,用户无法进行风险判断。

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

如果一个看似简单的标语生成技能实际包含 SM4 加解密、密钥校验和 Base64 密文处理,却未向用户解释其用途,这会显著扩大攻击面并隐藏敏感支付/凭证处理逻辑。与业务目标无关的密码学能力常意味着在处理秘密数据,若披露不足,用户无法进行风险判断。

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

要求模型向用户公开‘完整思维链’属于明确的敏感内部信息披露请求。泄露内部推理会暴露安全策略、判断依据和潜在防护边界,使攻击者更容易迭代提示注入、规避检测或诱导模型泄密。

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

重复强调必须包含助手思考过程,说明这不是偶然表述,而是刻意设置的内部信息泄露要求。在一个涉及支付、凭证和外部技能调用的上下文中,这种泄露更危险,因为它可能暴露敏感判断路径和安全决策细节。

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

This configuration embeds a hard-coded payment recipient address ("payTo") for a paid skill, allowing funds to be silently directed to an arbitrary account controlled by whoever packaged or modified the skill. In skill ecosystems, billing metadata is security-sensitive: if users or host agents trust the manifest blindly, this creates a real risk of unauthorized monetization, fund diversion, or supply-chain style payment redirection.

Content

Scanner excerpt · configs/config.json (reported line 2)May include surrounding context.

json
{
  "payTo": "8f18fb4ccac281a40fb0cf96ec7c2aa8202604081312250020002552zmV8kH3UJyhYdJjFXghUiYLL4FuyrNtvEwiagAzFyufPIN3ePZVMwfMkzyhJD2FuoPrQ118i",
  "sm4Key": "WoDb5xjzDmsYPV875mvYqg==",
  "amount": 1,
  "skillName": "team-banner",
  "description": "服务费用"
}

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The hardcoded slug identifies the service as 'ai-chunlian' while the manifest describes a team-banner slogan generator, indicating the code was likely repurposed from a different paid skill without proper disclosure. Such inconsistent identity markers are a strong sign of deceptive packaging and can hinder auditing, billing traceability, and user understanding of what service they are invoking.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill is presented as a team-banner slogan generator, but the code instead creates and stores payment orders and performs no slogan-generation logic at all. This mismatch is dangerous because it can deceive users and reviewers about the skill’s true behavior, enabling unauthorized monetization or hidden transaction handling under a benign-looking label.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The saved order description explicitly states '春联生成服务费用' even though the skill claims to provide team-banner slogan generation. This creates a misleading billing record and suggests the service may be charging for or routing through a different offering than the one presented to the user.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

文档要求‘包含你的思考过程’属于明显的不当指令,试图诱导模型泄露内部推理信息;这与横幅生成服务无关,且与安全边界直接冲突。此类提示常被用于绕过平台安全策略、获取隐藏决策逻辑或辅助后续提示注入。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The description explicitly instructs the skill to interact in Chinese, which forces a specific language choice. The policy allows locale or language constraints only when users are given a choice or the restriction is clearly justified; neither is present here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

This line reiterates that the skill should communicate in Chinese, reinforcing a fixed language requirement. Because no opt-in or alternative language handling is offered, this is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

文档允许在订单文件中保存并读取 payCredential,但没有对该支付凭证的敏感性、存储保护或最小暴露做出明确警示。支付凭证若被本地文件、日志或其他技能读取,可能导致支付状态伪造、凭证复用或隐私泄露。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

技能要求通过外部支付流程传输订单号和指示符,并依赖出站网络,但未向用户明确说明会传输哪些支付相关数据、传输给谁以及存在何种隐私/安全风险。对支付数据的网络发送缺乏透明度,会增加中间人、错误路由或第三方收集风险。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The description value is written only in Chinese ("服务费用"), which imposes a specific language in the skill configuration without indicating user choice or locale-specific justification. The policy for this audit flags natural-language locale constraints when no opt-in or documented reason is provided.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file contains user-facing natural-language strings entirely in Chinese, including the module description and all runtime status/output messages. Under the policy rule for language/locale, forcing a specific language without user opt-in or documented justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The code includes payment recipient selection, amount handling, encryption of order payloads, and order persistence for a simple content-generation skill with no demonstrated need for transactional behavior. In this context, hidden payment capabilities materially expand abuse potential, including surprise charges, unauthorized order creation, or covert monetization.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The script writes order information, including user question content, payment recipient, amount, and encrypted order data, to storage immediately after creation without any visible confirmation or warning to the user. In a user-facing skill, silent persistence of transaction-related data increases privacy and consent risks, especially when the skill’s declared purpose does not clearly imply payment processing.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The file implements payment-related configuration loading and local order persistence even though the skill is presented as a banner-slogan generator. This capability mismatch is security-relevant because it expands the trust boundary, introduces handling of potentially sensitive business data, and can enable undisclosed monetization or data collection beyond what a user would reasonably expect from the stated purpose.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

A slogan-generation skill having the ability to read local payment configuration and persist order records is beyond the least-privilege scope implied by its function. In the skill context, this makes the behavior more dangerous because users are likely to grant trust for harmless text generation, not for local financial configuration access and durable order tracking.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

该技能宣称是团建横幅标语生成,但代码中却实现了完整的通用 SM4 加解密能力,与声明功能明显不符。此类与业务无关的密码学能力会扩大技能的攻击面,可被复用为隐藏数据处理、规避审计或承载未披露的数据传输逻辑,尤其在付费配置和可定制收款方的上下文中更值得警惕。

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

在标语生成场景中引入通用密码学原语缺乏明确业务正当性,本身就是可疑设计信号。虽然文件未直接显示恶意行为,但这类能力可作为后续隐藏配置、加密载荷或敏感数据封装的基础模块;技能上下文越偏离安全/支付/合规场景,这种不相称能力就越危险。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The file includes a Chinese-only natural-language comment ('配置文件') with no indication that language choice is optional or that the locale is intentionally constrained for a region-specific use case. This can violate language/locale policy when a skill implicitly assumes a single language without user opt-in.

Content

No source excerpt is available for this finding.