The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a shell script from a remote URL to install the GEF (GDB Enhanced Features) plugin.
Evidence (SKILL.md): bash -c "$(curl -fsSL https://gef.blah.cat/sh)"
Evidence (references/binary-tools-reference.md): bash -c "$(curl -fsSL https://gef.blah.cat/sh)"
[COMMAND_EXECUTION]: The skill provides numerous examples of shell commands for binary analysis and exploitation, including debugger execution, network listeners, and system tracing.
Evidence (SKILL.md): Extensive list of commands including gdb, strace, ltrace, objdump, and qltool.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for cloning third-party repositories and downloading specific shellcodes from external sources.
Evidence (SKILL.md): git clone https://github.com/pwndbg/pwndbg and shellcode get 61 (via GEF).
[PRIVILEGE_ESCALATION]: The skill includes instructions to use sudo for tool installation and system configuration changes.
Evidence (SKILL.md): sudo apt-get install gdb and echo 0 | sudo tee /proc/sys/kernel/randomize_va_space (disabling ASLR).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze binary files which could potentially contain embedded instructions if processed by an LLM, though it lacks direct interpolation of untrusted strings into agent prompts.
Ingestion points: Processes local binary files (./binary), core dumps, and remote debugger streams.
Boundary markers: None present in the instructions.
Capability inventory: Extensive capabilities including shell execution (bash), file writes (dump binary), and network operations (remote('host', port)).
Sanitization: No sanitization of external tool outputs is mentioned.

binary-exploitation-tools