The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/main.cjs file uses child_process.spawn to run the AWS CLI tool with arguments provided by the agent. Although it implements shell: false to prevent shell injection, the skill acts as a general-purpose wrapper for the aws binary.
[EXTERNAL_DOWNLOADS]: The SKILL.md file directs users to download the AWS CLI v2 from official Amazon Web Services domains such as aws.amazon.com and awscli.amazonaws.com. These are established, trusted sources for cloud tooling.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it processes data from external cloud resources while maintaining the capability to execute commands. Evidence: 1. Ingestion points: AWS CLI command outputs entering the agent context via scripts/main.cjs. 2. Boundary markers: No explicit separators or 'ignore instructions' markers are defined in the wrapper scripts or documentation. 3. Capability inventory: Execution of arbitrary aws subcommands using child_process.spawn in scripts/main.cjs. 4. Sanitization: No specific input/output sanitization or validation logic is implemented in the provided Node.js scripts. Additionally, the SKILL.md file contains a claim that 'IAM write operations are blocked', which is a deceptive metadata instruction as it is not enforced by the underlying scripts/main.cjs implementation.

aws-cloud-ops