The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of compiled binaries, mobile applications, firmware blobs, and obfuscated scripts.
Ingestion points: The skill explicitly instructs the agent to process various binary artifacts and custom VM bytecodes during CTF tasks.
Capability inventory: The agent is provided with powerful diagnostic and manipulation tools including debuggers (GDB), disassemblers (Ghidra), dynamic instrumentation frameworks (Frida), and binary patching libraries (pwntools).
Boundary markers: The skill does not define specific delimiters for wrapping untrusted binary data or its decompiled/disassembled output within the agent's context, which is common in technical reference materials.
Sanitization: There is no specific instruction for the agent to sanitize or filter the results derived from analyzing these potentially adversarial artifacts.
[COMMAND_EXECUTION]: The skill provides a large variety of command-line examples for performing technical triage and binary analysis.
Evidence: Reference files like references/platforms.md and references/tools.md contain numerous commands such as otool, binwalk, checksec, readelf, and upx integrated into the workflow guides.
[DYNAMIC_EXECUTION]: The skill outlines workflows that involve runtime code generation, transpilation, and compilation for analysis purposes.
Evidence: references/tools.md describes converting WASM to C and then compiling it with gcc. references/patterns.md suggests using gcc to build optimized solvers for cryptographic challenges.

reverse-ctf