reverse-engineer-anything

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including binary files, archives, web page content, and network captures. • Ingestion points: Binary analysis tools, archive extraction, and browser scenario captures (SKILL.md, references/environments.md). • Boundary markers: Instructions advise citing stable Evidence IDs and noting coverage limitations, but do not provide explicit delimiters for untrusted content. • Capability inventory: High capabilities including shell command execution, file system writes via extraction, and interactive debugger sessions (references/runtime-observation.md). • Sanitization: The skill employs structured analysis views to filter and aggregate findings, reducing direct exposure to raw artifact data.
  • [REMOTE_CODE_EXECUTION]: The skill utilizes npx -y rea-agents@latest for CLI fallback operations and setup tasks, which downloads and executes code from the npm registry (SKILL.md).
  • [EXTERNAL_DOWNLOADS]: Fetches the rea-agents package and potential dependencies from the npm registry, which is a well-known service but involves unversioned package execution via the @latest tag.
  • [COMMAND_EXECUTION]: The skill executes various system commands for environment diagnostics and MCP server registration, such as rea setup and rea doctor (references/connection-and-recovery.md).
  • [DYNAMIC_EXECUTION]: The skill facilitates dynamic instrumentation and debugging through Frida, GDB, and Rizin, which involve executing code within target processes (references/runtime-observation.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 06:50 PM