aws-transform

Fail

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute an installation script from https://transform-cli.awsstatic.com/install.sh via a shell pipe. This source is an official AWS domain used by the skill author (awslabs) to distribute required tooling.
  • [PROMPT_INJECTION]: Instructions in SKILL.md and references/vmware-network.md command the agent to perform internal state checks and polling tasks silently, explicitly stating that the user should not see these operations. While this hides agent logic from the user interface, it is used here for UX management of background processes.
  • [DYNAMIC_EXECUTION]: To handle long-running transformations, the skill generates local bash scripts (e.g., run.sh, run-<repo-name>.sh) at runtime. These scripts are used to execute the CLI in the background and capture status codes.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface as it is designed to scan and process external source code from repositories (GitHub, GitLab, local). This ingestion of untrusted data combined with the skill's high-privilege capabilities (file writing, command execution) constitutes a standard operational risk for code-analysis agents.
  • [COMMAND_EXECUTION]: The skill utilizes shell-based file editing via perl, sed, and powershell to modify user DDL/SQL files. It includes robust security guidance against echoing secrets or tokens in chat and process history.
  • [EXTERNAL_DOWNLOADS]: Dependencies such as the AWS Transform MCP server and the AWS CDK are installed from official package registries at runtime.
Recommendations
  • HIGH: Downloads and executes remote code from: https://transform-cli.awsstatic.com/install.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 5, 2026, 09:19 AM