The Agent Skills Directory

[SAFE]: The skill defines a structured workflow for performing security assessments of native binaries such as EXEs, DLLs, and ELF files.
[SAFE]: It incorporates safety best practices by explicitly instructing the agent to execute suspicious code only in authorized, isolated environments.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted native binaries which could theoretically contain instructions intended for the agent. However, the methodology emphasizes isolation, verification, and fact-checking, which are effective mitigations for this domain.
Ingestion points: Target binaries (EXE, DLL, ELF, Mach-O, firmware images).
Boundary markers: Uses a 'scope ledger' and structured slash commands to maintain assessment boundaries.
Capability inventory: Uses binary analysis tools via commands like /capa and /ghidra.
Sanitization: Instructions mandate the use of isolated environments for dynamic analysis.

native-binary-assessment