The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use sudo for high-privilege system modifications. This includes disabling Address Space Layout Randomization (ASLR) on Linux via echo 0 | sudo tee /proc/sys/kernel/randomize_va_space and modifying the macOS system keychain to trust external certificates using sudo security add-trusted-cert.
[COMMAND_EXECUTION]: Extensive use of shell commands for process management and environment cleanup, including pkill, rm, lsof, and git checkout. It also uses powerful tracing tools like strace, ltrace, and dtruss (via sudo).
[EXTERNAL_DOWNLOADS]: The skill directs the agent to download and install multiple external packages and tools from public registries and repositories, such as pwntools, playwright, ghidra, pyinstxtractor.py, and mitmproxy.
[REMOTE_CODE_EXECUTION]: Dynamically generates and executes Python scripts (e.g., extract_bundled_js.py) to carve and extract data from binary files at runtime. It also recommends running potentially untrusted binaries under debuggers like gdb and lldb.
[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted binary data and code, creating a surface for indirect prompt injection. While the methodology includes some verification steps, the high-capability environment (access to shell, network, and file system) increases the risk of exploitation if instructions are embedded in analyzed artifacts.

debugging