The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to use the Zig compiler (zig build-exe) to compile source code provided in the documentation (e.g., in references/examples.md). While intended for creating reference binaries for comparison, this involves runtime compilation of code snippets provided within the skill content.
[INDIRECT_PROMPT_INJECTION]: The skill establishs a vulnerability surface by instructing the agent to ingest and process untrusted data from external binary files using tools like readelf, llvm-nm, and strings. A maliciously crafted binary could potentially include strings or metadata designed to influence the agent's reasoning. * Ingestion points: Output from shell commands readelf, llvm-nm, llvm-objdump, strings, and zig processing external files (e.g., sample or hello_dbg) mentioned in SKILL.md and references/examples.md. * Boundary markers: No explicit instructions or delimiters are provided to the agent to distinguish between tool diagnostic output and potentially malicious strings embedded within the analyzed binaries. * Capability inventory: The skill utilizes shell execution for binary analysis tools and the Zig compiler. * Sanitization: No specific filtering or sanitization of the binary analysis tool output is described.

re-zig