prompt-library-for-startups
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [Command Execution Patterns]: The skill provides a library of system prompts designed to interact with the AWS CLI, Docker, and other developer tools to perform tasks like infrastructure auditing, deployment, and security scanning. While these involve executing terminal commands, the skill explicitly instructs the agent to display every command for user review and wait for confirmation before execution.
- [External Resource References]: The skill includes links to 'Downloadable Agents' hosted on the
aws-samplesGitHub organization. These resources are from a trusted vendor and provide automation for cost monitoring and quota management. They are documented neutrally and require separate installation by the user. - [Privilege Management Guidance]: Several prompts, particularly for initial account setup and cost auditing, guide the user in creating IAM roles and policies. These instructions emphasize the principle of least privilege, such as generating a 'ViewOnlyAccess' role for audits, and provide restrictive JSON policy templates to minimize risk.
- [Indirect Prompt Injection Surface]: Prompts designed for auditing and code generation ingest data from external sources like AWS API responses or local source files. To address the risk of embedded instructions in this data, the skill includes explicit 'Operating Rules' that mandate treating all API-returned text as untrusted and dropping sensitive fields (such as environment variables or credentials) immediately upon receipt.
- [Data Exposure Protection]: The 'Cost Optimization Audit' prompt implements strict data-handling rules (Rule 4) that require the agent to mask sensitive identifiers, such as Account IDs and ARNs, in all outputs. It also includes instructions for managing a 'workpaper' file with restrictive permissions to prevent accidental exposure of sensitive spend data.
Audit Metadata