The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to use npx -y rea-agents@latest to download and execute code from the npm registry at runtime, which can lead to arbitrary code execution if the package is compromised.
[EXTERNAL_DOWNLOADS]: Guides the user and agent to fetch tools, configuration, and documentation from untrusted third-party repositories, such as github.com/morluto/rea and github.com/mrexodia/ida-pro-mcp.
[PERSISTENCE]: The rea setup command is used to modify local client configuration files (such as mcp.json for Codex or Claude Desktop), which registers the analysis server to persist across agent sessions.
[COMMAND_EXECUTION]: Orchestrates the execution of multiple high-privilege system utilities, including debuggers (gdb, lldb), network traffic analyzers (tshark), and decompilation tools (jadx), allowing significant interaction with the host system.
[INDIRECT_PROMPT_INJECTION]: Exhibits a wide vulnerability surface by processing untrusted data formats—such as firmware images, Android APKs, and network captures—without explicit sanitization or boundary markers to prevent the data from influencing agent behavior.
[PRIVILEGE_ESCALATION]: The skill uses observe_native_calls to launch targets under the lldb debugger and rea setup to write to application-specific configuration directories, both of which require significant host permissions.
[DATA_EXFILTRATION]: Provides capabilities to read and analyze sensitive local data, including network captures (HAR/SAZ/PCAP), SQLite database snapshots, and recorded crash dumps, creating a potential path for data exposure.

reverse-engineer-anything