The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from Alibaba Cloud API responses, such as instance lists and attributes, and has the capability to execute high-impact commands like delete-db-instance and reset-account-password. This ingestion of external data alongside destructive capabilities defines a vulnerability surface for indirect prompt injection, where malicious content in resource names or metadata could attempt to influence the agent's actions.
Ingestion points: Data from aliyun dds describe-db-instances and describe-db-instance-attribute processed in SKILL.md and references/operations.md.
Boundary markers: Mandatory human-in-the-loop confirmation for all write and delete operations, including a requirement for the user to type a specific confirmation string for deletions.
Capability inventory: Commands for instance deletion, password resetting, and firewall modification available across multiple management scripts.
Sanitization: No explicit programmatic sanitization of API data is performed; the skill relies on displaying parameters for user review before execution.
[EXTERNAL_DOWNLOADS]: The CLI installation guide within the skill references official download links for aliyun-cli from aliyuncli.alicdn.com. These are verified vendor resources necessary for the skill's operation.
[COMMAND_EXECUTION]: The skill uses the aliyun CLI as its primary mechanism for interacting with cloud infrastructure. All commands are consistent with the skill's purpose of managing MongoDB resources.
[PRIVILEGE_ESCALATION]: The installation instructions in references/cli-installation-guide.md mention the use of sudo for installing system packages on Linux distributions. This is typical for administrative tool installation and is not considered malicious in this context.

alibabacloud-mongodb-instances-manage