The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of firmware images (firmware.bin) and extracted file systems.
Ingestion points: Firmware files downloaded from external URLs or extracted via hardware interfaces (UART, JTAG).
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious embedded content within the analyzed firmware files.
Capability inventory: The skill leverages powerful system tools including dd, screen, sudo chroot, gcc cross-compilers, and qemu emulation.
Sanitization: There is no evidence of sanitization or validation of the content extracted from firmware before it is processed by analysis tools.
[PRIVILEGE_ESCALATION]: The instructions include usage of sudo chroot to emulate firmware environments. While this is a standard procedure for firmware analysis to provide a consistent root filesystem, it involves executing commands with administrative privileges.
[COMMAND_EXECUTION]: The skill directs the agent to use low-level system utilities such as dd for partition dumping and screen for serial console access. It also includes examples of vulnerable C code patterns (system(cmd)) for educational purposes to demonstrate command injection risks in target firmware.
[DYNAMIC_EXECUTION]: The workflow includes cross-compiling source code using arm-linux-gnueabi-gcc and mipsel-linux-gnu-gcc to create test binaries/exploits, as well as executing arbitrary binaries within a QEMU-emulated environment.

firmware-analyst