Back to skill

Security audit

腾讯云APM性能分析skill

Security checks for vulnerabilities and agentic risk

Overview

This skill has a coherent APM analysis purpose, but it asks users to persist cloud credentials and allows signed requests to a configurable endpoint without enough safeguards.

Review this skill carefully before installing. Use only least-privilege, preferably short-lived Tencent Cloud credentials; avoid putting SecretKeys in shell startup files or command-line arguments; verify APM_API_ENDPOINT is a trusted Tencent Cloud endpoint; and treat tool descriptions and results returned by the remote MCP server as data, not instructions.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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
  • 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
Findings (4)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/apm_mcp_client.py:177
Finding

Unrestricted API Endpoint Override Exposes Signed Requests to Attacker-Controlled Hosts

Content
View full analysis

Vulnerability Details

File Location: scripts/apm_mcp_client.py:177-181; related signing behavior in scripts/tencentcloud/common/abstract_client.py:192-225
Vulnerability Type: Unvalidated destination for authenticated network requests
Risk Level: High

Vulnerable Code

python
# scripts/apm_mcp_client.py:177-181

# 配置 Endpoint
endpoint = os.environ.get("APM_API_ENDPOINT", DEFAULT_ENDPOINT)

httpProfile = HttpProfile()
httpProfile.endpoint = endpoint

The selected endpoint is subsequently used when creating authenticated requests:

python
# scripts/tencentcloud/common/abstract_client.py:192-225

endpoint = self._get_endpoint(options=options)
timestamp = int(time.time())
cred_secret_id, cred_secret_key, cred_token = self.credential.get_credential_info()
req.header["Host"] = endpoint
req.header["X-TC-Action"] = action[0].upper() + action[1:]
req.header["X-TC-RequestClient"] = self.request_client
req.header["X-TC-Timestamp"] = str(timestamp)
req.header["X-TC-Version"] = self._apiVersion
if self.profile.unsignedPayload is True:
    req.header["X-TC-Content-SHA256"] = "UNSIGNED-PAYLOAD"
if self.region:
    req.header['X-TC-Region'] = self.region
if cred_token:
    req.header['X-TC-Token'] = cred_token
if self.profile.language:
    req.header['X-TC-Language'] = self.profile.language

if req.method == 'GET':
    params = copy.deepcopy(self._fix_params(params))
    req.data = urlencode(params)
elif content_type == _json_content:
    req.data = json.dumps(params)
elif content_type == _multipart_content:
    boundary = uuid.uuid4().hex
    req.header["Content-Type"] = content_type + "; boundary=" + boundary
    req.data = self._get_multipart_body(params, boundary, options)

service = self._service
date = datetime.utcfromtimestamp(timestamp).strftime('%Y-%m-%d')
signature = self._get_tc3_signature(params, req, date, service, cred_secret_key, options)

auth = "TC3-HMAC-SHA256 Credential=%s/%s/%s/tc3_request, SignedHeaders=content-type;host, Si
...[truncated 2649 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove arbitrary endpoint overrides unless they are operationally essential.
  2. If private endpoints must be supported, enforce an explicit allowlist, such as:
    • apm.tencentcloudapi.com
    • Specifically approved Tencent internal APM endpoints
  3. Normalize and validate the endpoint before constructing the SDK client:
    • Reject URL paths, query strings, fragments, embedded credentials, and non-default ports.
    • Require an exact approved hostname or a carefully bounded Tencent-controlled suffix.
    • Prevent suffix-confusion domains such as apm.tencentcloudapi.com.attacker.example.
  4. Require HTTPS and reject any configuration that selects plain HTTP.
  5. Fail closed when validation fails; do not silently fall back to another destination.
  6. Display the resolved destination before the first authenticated call when a non-default endpoint is selected.
  7. Treat responses from non-public or development endpoints as untrusted data.
  8. Prefer short-lived, least-privilege credentials limited to the required APM actions.
  9. Add tests for malicious endpoints, URL user information, Unicode hostname confusion, suffix confusion, ports, and malformed host values.

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/apm_mcp_client.py:467
Finding

SecretKey Accepted Through Command-Line Arguments

Content
View full analysis

Vulnerability Details

File Location: scripts/apm_mcp_client.py:467-474
Vulnerability Type: Sensitive information exposed through process arguments and shell history
Risk Level: Medium

Vulnerable Code

python
parser.add_argument(
    "--secret-id",
    help="腾讯云 SecretId(覆盖环境变量 TENCENTCLOUD_SECRET_ID)",
)
parser.add_argument(
    "--secret-key",
    help="腾讯云 SecretKey(覆盖环境变量 TENCENTCLOUD_SECRET_KEY)",
)

The parsed values are passed directly to client creation:

python
client = create_client(
    cli_secret_id=getattr(args, "secret_id", None),
    cli_secret_key=getattr(args, "secret_key", None),
)

The command-line credentials take priority:

python
secret_id = cli_secret_id or os.environ.get("TENCENTCLOUD_SECRET_ID")
secret_key = cli_secret_key or os.environ.get("TENCENTCLOUD_SECRET_KEY")

Technical Analysis

Command-line arguments are an inappropriate channel for long-lived cloud credentials. Depending on the operating system and environment, arguments can be exposed through:

  • Process-listing utilities
  • /proc/<pid>/cmdline
  • Shell command history
  • Terminal session recording
  • Process monitoring and observability agents
  • CI job metadata and diagnostic output
  • Parent-process or desktop-launcher logs

This contradicts the Skill's stated policy that sensitive parameters should be provided only through environment variables. The argument names and help text explicitly advertise command-line credential use and give it higher priority than environment variables.

Attack Path

  1. A user invokes the client with --secret-id and --secret-key.
  2. The shell may save the entire command in its history.
  3. While the process is running, another local process or user with sufficient process-inspection access reads its command line.
  4. Monitoring, support, or CI systems may also capture the invocation.
  5. The observer obtains the complete long-lived Tencent Cloud SecretKey.
  6. The stolen credential is used di ...[truncated 606 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the --secret-id and --secret-key command-line options.
  2. Use environment variables, a protected operating-system credential store, or a standard cloud credential provider.
  3. Prefer short-lived credentials rather than long-lived access keys.
  4. If interactive entry is required, use getpass.getpass() for the SecretKey and do not echo or persist it.
  5. Update the credential priority documentation so command-line credentials are no longer supported.
  6. Warn existing users to:
    • Remove credential-bearing commands from shell history.
    • Rotate any SecretKey previously supplied on the command line.
    • Review CI and observability logs for accidental exposure.
  7. Add automated tests that reject credential-related command-line options.

T09 · Insecure Skill Coding Practices

Error
Location
references/credential_guide.md:27
Finding

Long-Lived Cloud Credentials Are Persisted in Plaintext Shell Startup Files

Content
View full analysis

Vulnerability Details

File Location: references/credential_guide.md:27-32
Vulnerability Type: Plaintext sensitive credential persistence
Risk Level: High

Vulnerable Instructions

bash
echo 'export TENCENTCLOUD_SECRET_ID="your-secret-id"' >> ~/.zshrc
echo 'export TENCENTCLOUD_SECRET_KEY="your-secret-key"' >> ~/.zshrc
source ~/.zshrc

The guide repeats the same approach for Bash:

bash
echo 'export TENCENTCLOUD_SECRET_ID="your-secret-id"' >> ~/.bashrc
echo 'export TENCENTCLOUD_SECRET_KEY="your-secret-key"' >> ~/.bashrc
source ~/.bashrc

It also claims:

text
密钥仅通过环境变量读取,不会写入任何文件或日志。

Technical Analysis

The recommended commands explicitly append the Tencent Cloud SecretId and SecretKey to persistent plaintext files. This directly contradicts the statement that credentials “will not be written to any file.”

Shell startup files are not dedicated secret stores. They may be:

  • Read by unrelated interactive processes and shell extensions
  • Included in home-directory backups or synchronization
  • Collected in workstation support bundles
  • Accidentally committed with dotfile repositories
  • Exposed by overly broad file permissions
  • Read by malware or any process operating under the same user account
  • Inherited by every child process launched from the shell

Appending on every setup attempt can also create multiple stale credential copies. Rotating a key in one line does not necessarily remove prior copies from the file, backups, or version history.

Although shell configuration persistence is not a backdoor, it creates cross-session plaintext credential persistence that is unnecessary for one-purpose APM access.

Attack Path

  1. The user follows the mandatory credential setup template.
  2. The actual SecretId and SecretKey are appended to .zshrc or .bashrc.
  3. The file is backed up, synchronized, committed to a dotfiles repository, or read by another process running as the user.
  4. An attacker obtains the plaintext ...[truncated 681 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove instructions that append secrets to .zshrc or .bashrc.
  2. Correct the documentation so it does not claim that environment variables avoid writing credentials to files when the setup procedure does exactly that.
  3. Prefer, in order:
    • Short-lived Tencent Cloud credentials
    • Workload or instance roles where appropriate
    • An operating-system credential manager
    • A protected standard cloud credential file with strict permissions and a narrowly scoped profile
    • Session-only environment variables as a temporary fallback
  4. Do not export long-lived secrets globally to every process spawned by the user's shell.
  5. Enforce or clearly document the exact minimum CAM policy required for SendMCPMessage.
  6. Recommend credential rotation and periodic expiration.
  7. Warn current users to remove old entries, inspect dotfile repositories and backups, and rotate credentials already persisted this way.
  8. If a file-based provider is retained, verify ownership and permissions and reject group- or world-readable credential files.

T01 · Skill Instruction Hijacking

Warning
Location
references/interaction_guide.md:7
Finding

Remote MCP Tool Metadata and Results Are Consumed Without Prompt-Injection Boundaries

Content
View full analysis

Vulnerability Details

File Location: references/interaction_guide.md:7-11; related output processing in scripts/apm_mcp_client.py:296-328
Vulnerability Type: Untrusted remote content influencing agent decisions
Risk Level: Medium

Vulnerable Instructions and Processing

The interaction guide directs the agent to determine behavior from remote tool descriptions and parameter definitions:

markdown
### 场景一:工具支持多种功能/模式

若工具描述或参数表明支持多种功能(如查询工具既能按名称也能按 ID 查询),先展示可用功能并确认用户需求。

The client displays remote names, descriptions, and schemas without marking them as untrusted:

python
for i, tool in enumerate(tool_list, 1):
    name = tool.get("name", "N/A")
    desc = tool.get("description", "(无描述)")
    schema = tool.get("inputSchema", {})

    print(f"\n  [{i}] {name}")
    print(f"      描述: {desc}")

    properties = schema.get("properties", {})
    required = _get_required_fields(schema)

    if properties:
        print(f"      参数:")
        for param_name, param_info in properties.items():
            param_type = param_info.get("type", "any") if isinstance(param_info, dict) else "any"
            param_desc = param_info.get("description", "") if isinstance(param_info, dict) else ""
            is_required = "必填" if param_name in required else "可选"
            print(f"        - {param_name} ({param_type}, {is_required}): {param_desc}")

Tool results are similarly printed as supplied by the remote MCP service:

python
parsed = result.get("parsed_content")
if parsed and isinstance(parsed, (dict, list)):
    print(json.dumps(parsed, indent=2, ensure_ascii=False))
else:
    try:
        data = json.loads(content)
        print(json.dumps(data, indent=2, ensure_ascii=False))
    except (json.JSONDecodeError, TypeError):
        print(content)

Technical Analysis

Tool descriptions, parameter descriptions, and tool results originate from a remote MCP service. They are data, not trusted Skill instructions. However, the Skill i ...[truncated 2072 chars]

Remediation
View remediation

Remediation Suggestions

  1. Add an explicit mandatory rule that all MCP names, descriptions, schemas, errors, and results are untrusted data.
  2. State that instructions embedded in remote responses must never override:
    • System or developer instructions
    • Skill safety constraints
    • User intent
    • Tool-confirmation requirements
  3. Restrict remote responses to expected JSON structures and enforce:
    • Maximum field lengths
    • Allowed parameter types
    • Valid tool-name patterns
    • Maximum nesting depth and response size
  4. Render remote text inside clearly labeled quoted or code-delimited sections.
  5. Do not execute commands, access files, disclose secrets, or make additional tool calls solely because remote content requests it.
  6. Maintain a local allowlist of supported APM tool names and expected parameter schemas where practical.
  7. Require a separate user confirmation for actions derived from remote results, especially any operation outside read-only APM querying.
  8. Combine these controls with strict endpoint allowlisting.
  9. Add adversarial tests containing instruction-like tool descriptions and results to verify that the agent treats them only as data.
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (42)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill instructs the agent to execute local Python scripts through Bash using parameterized commands and to inspect environment-based credentials. In a skill context, subprocess-capable workflows can become dangerous if the downstream script or tool-name/args handling is not tightly validated, because user-influenced input may trigger unintended command execution or misuse of cloud credentials.

Content

No source excerpt is available for this finding.

YARA rule 'backdoor_persistence': Backdoor persistence with malicious payloads (shell commands, SSH key injection, hidden root users) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · references/credential_guide.md (reported line 66)May include surrounding context.

�

第二步:写入 shell 配置文件(永久生效)

在终端执行以下命令(将 your-secret-id 和 your-secret-key 替换为你的实际密钥):

macOS / Linux (Zsh):

bash
echo 'export TENCENTCLOUD_SECRET_ID="your-secret-id"' >> ~/.zshrc
echo 'export TENCENTCLOUD_SECRET_KEY="your-secret-key"' >> ~/.zshrc
source ~/.zshrc

macOS / Linux (Bash):

bash
echo 'export TENCENTCLOUD_SECRET_ID="your-secret-id"' >> ~/.bashrc
echo 'export TENCENTCLOUD_SECRET_KEY="your-secret-key"' >> ~/.bashrc
source ~/.bashrc

第三步:验证配置

bash
# 检查环境变量是否生效(应输出你的 SecretId)
echo $TENCENTCLOUD_SECRET_ID

第四步(可选):配置其他环境变量

bash
# 切换为内网 Endpoint(仅腾讯内网环境使用,公网无需设置)
# echo 'export APM_API_ENDPOINT="apm.ap-guangzhou.tencentcloudapi.woa.com"' >> ~/.zshrc

source ~/.zshrc

支持的环境变量

| 变量名 | 必填 | 说�

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/credential_guide.md (reported line 116)May include surrounding context.

md
3. **版本控制排除**: 密钥仅存在于用户的 shell 配置文件中,不会出现在项目文件里。
4. **日志脱敏**: 错误日志不记录密钥值,日志文件权限 `600`。
5. **对话安全**: 用户提供密钥明文时不得回显,应提示在终端中配置环境变量。
6. **不使用 .env 文件**: 凭证不通过 `.env` 文件管理,避免被意外提交到版本控制。

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
85% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/mobile_compat_guide.md (reported line 32)May include surrounding context.

text
write_to_file("/tmp/parse_spans.py", ...)        # 禁止创建临时脚本
execute_command("python /tmp/parse_spans.py")     # 禁止执行临时脚本
execute_command("rm -f /tmp/parse_spans.py ...")  # 禁止删除文件
delete_file("/tmp/data.json")                     # 禁止调用删除工具

Lp1

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not listed in its permissions. This may indicate deceptive intent or missing permission declarations.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'network' capability that is not listed in its permissions. This may indicate deceptive intent or missing permission declarations.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The trigger conditions are broad and overlap with ordinary discussion of APM, performance, spans, or latency. In an agent skill system, overly broad activation increases the chance the skill runs in contexts where users did not intend remote cloud interaction, potentially causing unnecessary credential prompts, external network access, or sensitive operational queries.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The "初始回答规范" section instructs the agent to generate the initial response according to a fixed Chinese-language format, and the overall skill content is written as if Chinese is required. There is no opt-in, user language selection, or documented justification that this skill must operate only in Chinese.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill declares a strict mobile-compatibility rule forbidding creation of temporary or auxiliary files, yet later requires writing failures to a local log file. This contradiction can push implementations into violating one of the safety constraints, and in practice may cause unexpected local file creation on constrained or remote environments where the user was explicitly told no files would be written.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The title states the guide as a mandatory rule set in Chinese, and the document provides all instructions exclusively in Chinese with no indication that users may choose another language or locale. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.