reverse-engineer-anything
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a comprehensive instruction manual for the REA (Reverse Engineering Assistant) framework. It includes specific workflows for different platforms (Android, .NET, Electron, Native).
- [REMOTE_CODE_EXECUTION]: The skill mentions using
npx -y rea-agents@latestfor diagnostics and setup. While this executes remote code from npm, it is a standard deployment pattern for CLI tools and is explicitly scoped to the user's local environment with dry-run options and approval checkpoints. The skill author 'zouineoualid' is consistent with the 'rea-agents' ecosystem. - [DATA_EXFILTRATION]: No patterns of data exfiltration were detected. The skill explicitly notes that sensitive data like credentials, cookies, and authorization headers are 'deliberately absent' from its runtime observation tools.
- [COMMAND_EXECUTION]: The skill utilizes standard CLI commands for its purpose. It instructs the agent to obtain explicit approval before applying configuration changes (
--yes) or installing secondary engines like Hopper. This demonstrates a high regard for user control. - [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data (binary/package contents). It mitigates risk by specifying that analysis is static (non-executing) and by instructing the agent to 'distinguish observations, inferences, and unknowns,' which prevents the agent from blindly obeying instructions that might be embedded in analyzed files.
- [PRIVILEGE_ESCALATION]: The skill does not request sudo or administrative privileges. It uses user-scoped directories for temporary extraction and standard MCP (Model Context Protocol) registration paths.
Audit Metadata