The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external, potentially untrusted binaries through string searches, memory dumps, and disassembly views. This creates a surface where embedded malicious instructions could influence agent behavior. * Ingestion points: The agent reads output from the debugger, including strings (Shift+D), memory contents (Ctrl+E, savedata), and disassembly results. * Boundary markers: The instructions do not define specific delimiters to distinguish between debugger output and agent commands. * Capability inventory: The skill utilizes debugging operations (attach, single-step, breakpoint), file system writes (memory dumps, IAT fixing), and process manipulation. * Sanitization: There are no explicit steps for sanitizing or validating the strings or data retrieved from the target process.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading tools from public repositories and established websites. * Evidence: References downloading x64dbg from the x64dbg/x64dbg GitHub repository, Scylla from the NtQuery/Scylla GitHub repository, and OllyDbg from the ollydbg.de domain.

re-x64dbg