The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill describes an "On-Demand Bootstrap" system that automatically downloads and installs the Radare2 environment from the official GitHub releases if the tools are not detected on the host system.
[EXTERNAL_DOWNLOADS]: Fetches software packages and toolchain binaries from the Radare2 GitHub repository (radareorg/radare2) to establish the necessary execution environment.
[DYNAMIC_EXECUTION]: Prompts the agent to use the r2pm plugin manager to dynamically install and load functional extensions such as r2ghidra or r2dec during runtime.
[COMMAND_EXECUTION]: Extensively utilizes PowerShell and the Radare2 suite of command-line tools to interact with the file system, perform binary reconnaissance, and conduct file patching.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is the analysis of untrusted binary data (such as EXE, DLL, or ELF files), which constitutes a potential attack surface where instructions embedded in file strings or metadata could attempt to influence the agent's logic.
Ingestion points: Binary files processed through tools like rabin2 and r2.
Boundary markers: The skill mandates the use of specific "Evidence" identifiers (e.g., E-imports, E-exports) and checklists, which serve as structured manual verification steps for the agent.
Capability inventory: Includes shell command execution (PowerShell), binary file modification (via write mode r2 -w), and local network interaction (via r2http).
Sanitization: The provided documentation does not specify sanitization or escaping protocols for strings or data extracted from untrusted binaries.

radare2