webapp-pentesting
Fail
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The 'Authorization Status' section in the skill instructions contains explicit directives to override safety guidelines and ethical constraints, instructing the agent to omit disclaimers and ignore safety protocols during execution.
- [COMMAND_EXECUTION]: The skill implements a shell-based detection mechanism for a root directory ($SUPERHACKERS_ROOT) that includes searching within the user's home directory and plugin caches, subsequently sourcing and executing scripts from these potentially untrusted locations.
- [REMOTE_CODE_EXECUTION]: The skill utilizes 'eval' commands on the output of sourced scripts and provides browser automation code (Playwright/Puppeteer) configured to disable essential security features such as 'web-security' and the browser 'sandbox'.
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of multiple third-party tools and libraries from public package registries (Go, NPM, PyPI) and GitHub repositories without specifying versions or implementing integrity checks, which introduces supply chain vulnerabilities.
Recommendations
- HIGH: Downloads and executes remote code from: https://TARGET/graphql - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata