The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted binary files, creating an inherent attack surface for processing potentially malicious data.
Ingestion points: The bundled scripts/triage.py reads user-provided binary files; the agent is instructed to extract and interpret strings, headers, and metadata from unknown executables.
Boundary markers: SKILL.md enforces a 'mandatory execution gate' requiring confirmation of a dedicated VM, snapshot restoration capabilities, and network isolation before proceeding with analysis.
Capability inventory: The skill utilizes shell execution, debugger control (GDB, x64dbg, radare2), and network capture tools within the defined analysis environment.
Sanitization: The skill relies on strict environmental isolation (sandboxing/virtualization) rather than content sanitization, which is appropriate for the reverse engineering domain.
[COMMAND_EXECUTION]: The skill instructs the agent to use various command-line tools for binary inspection and analysis, such as file, strings, binwalk, readelf, and objdump.
[PRIVILEGE_ESCALATION]: The references mention sudo for specific tasks like chroot or tcpdump, but these instructions are scoped explicitly to a verified, isolated analysis virtual machine to prevent host compromise.

reverse-engineering