T05 · Unauthorized Access and Privilege Escalation
- Location
SKILL.md:166- Finding
Overly Broad Wildcard Permissions for Tencent Cloud Resources
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
The skill appears intended for Tencent Cloud server administration, but it asks users to grant broad cloud powers and includes unsafe credential and destructive-operation guidance.
Review this carefully before installing. Use a dedicated low-privilege Tencent Cloud sub-user, replace wildcard permissions with a narrow allowlist, never print or share config/.env, install dependencies in a pinned virtual environment, and require explicit human approval before creating prepaid resources or deleting instances.
SKILL.md:166Overly Broad Wildcard Permissions for Tencent Cloud Resources
SKILL.md:270Unpinned Dependencies Installed into the System Python Environment
SKILL.md:460Troubleshooting Instructions Expose Plaintext Cloud Credentials
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
```bash
# 复制示例文件
cd skills/tencentcloud-cvm
cp config/.env.example config/.env
# 编辑 .env 文件
vim config/.env
The instruction to inspect secrets by running cat config/.env directly exposes active Tencent Cloud credentials in plaintext. In this skill context, those credentials can control billable cloud resources, making accidental disclosure especially dangerous.
# 检查 .env 文件
cat config/.env
# 验证密钥
This finding is part of the same troubleshooting flow that encourages unsafe secret handling around .env verification. In the context of cloud administration, any workflow that normalizes printing or manually inspecting secrets increases the likelihood of credential leakage to logs, transcripts, or third parties.
# 检查 .env 文件
cat config/.env
# 验证密钥
python3 src/verify_config.py
The terminate_instance method deletes cloud resources immediately with no confirmation prompt, dependency check, or protection against accidental invocation. In an agent-operated environment, this materially increases the risk of irreversible service disruption and data loss from user error, automation mistakes, or adversarial prompt influence.
The skill advertises destructive operations such as creating and deleting cloud servers without an explicit warning about service interruption, data loss, billing consequences, or irreversibility. In an infrastructure-management context, omission of such warnings increases the chance of unsafe execution by users or downstream agents.
The skill states CAM permissions are not granted, yet the setup workflow instructs users to perform CAM user and access-key management actions. This contradictory guidance can mislead users about the true privilege requirements and encourage unsafe bootstrap practices using elevated identities.
The documentation claims the skill follows minimum 'CVM only' permissions, but the example policy grants wildcard access to CVM, VPC, and CBS resources. This mismatch can cause operators to deploy broader privileges than intended, increasing blast radius if the credentials are misused or the automation is compromised.
The troubleshooting section tells users to display the full .env file with cat config/.env, which can expose cloud API secrets in terminal history, logs, screen shares, or agent transcripts. Because these are active Tencent Cloud credentials, disclosure could enable unauthorized infrastructure access and resource abuse.
The create_instance method provisions billable cloud resources immediately, including prepaid instances, without any confirmation gate, dry-run mode, cost warning, or policy check. In an agent skill context, this is risky because a mistaken prompt, prompt injection, or misuse of the API wrapper can trigger real financial impact and infrastructure creation.
At L554, the docstring says KEEP_CHARGING means '停机不收费' and NOT_CHARGING means '停机收费', which contradicts the obvious semantics of the parameter names and the code's use of the StoppedMode field. This can mislead operators into choosing a shutdown mode with the opposite billing effect from what the documentation claims.