Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed, read-only Alibaba Cloud Bastion Host diagnostic helper, with cloud access that is sensitive but aligned with its purpose.
Before installing, ensure the agent will use a least-privilege Alibaba Cloud profile with only the listed read-only Bastion Host and Cloud Firewall permissions. Review each displayed aliyun command before allowing execution, because the results may reveal instance metadata, endpoints, firewall protection state, and traffic-log details.
The skill advertises activation for broad symptoms even when the user does not mention Bastion Host, which increases the chance the agent runs cloud-enumeration logic in unrelated contexts. That can lead to unnecessary access to cloud metadata and diagnostic APIs based on weak intent matching, violating least-privilege and potentially exposing infrastructure details.
The skill permits autonomous execution of CLI-based cloud queries after user authorization, which still creates a meaningful risk because the workflow can enumerate instances, endpoints, firewall assets, and traffic logs. In a mis-triggered or over-broad invocation, this turns the agent into an automated reconnaissance tool against the user's cloud environment.
- **Only perform read-only queries; strictly prohibit any write, modify, or delete operations**
- All Alibaba Cloud API calls must use `aliyun-cli` commands
- After the user explicitly authorizes automatic execution, the Agent may automatically execute the read-only `aliyun-cli` query commands listed in this Skill
- Before executing each command, display the command to be executed and explain its purpose
- If `aliyun-cli` is unavailable, credentials are not configured, or permissions are insufficient, stop automatic execution and inform the user of manual troubleshooting steps
- In uncertain situations, prioritize asking the user for additional information; do not perform write operations on your own
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
- After the user explicitly authorizes automatic execution, the Agent may automatically execute the read-only `aliyun-cli` query commands listed in this Skill
- Before executing each command, display the command to be executed and explain its purpose
- If `aliyun-cli` is unavailable, credentials are not configured, or permissions are insufficient, stop automatic execution and inform the user of manual troubleshooting steps
- In uncertain situations, prioritize asking the user for additional information; do not perform write operations on your own
- **Never expose or persist credentials**: do not copy, back up, move, print, `cat`, or read the aliyun CLI config/credential file (`~/.aliyun/config.json`) or any file containing an AccessKey, AccessKeySecret, or STS token, and never write credentials to any file. Credentials are pre-configured by the runtime environment; do not modify or back them up.
- **Run every `aliyun-cli` command as a SINGLE LINE** without backslash (`\`) line continuations, so that the command and all of its parameters appear on one physical line.
No suspicious patterns detected.