The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and model logic derived from third-party reverse engineering tools like Ghidra or IDA Pro.
Ingestion points: Decompiled pseudo-code, comparison chains, and mathematical relations extracted from potentially malicious binary files enter the agent's context.
Capability inventory: The instructions involve the creation and execution of Python scripts using the Z3 library to derive values like flags or keys.
Boundary markers: The skill does not define specific delimiters for separating raw decompiled input from instructions, leaving an attack surface for instructions hidden within code comments or logic.
Sanitization: The skill mandates the use of a sandbox environment for verifying solver outputs, providing a significant mitigation against malicious code execution resulting from solver results.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the Z3 solver and related dependencies.
Evidence: Recommends using pip install z3-solver, apt install z3, brew install python, and choco install python.
Context: These operations utilize official package managers and target well-known, reputable software registries, which are considered standard and safe practices.

re-z3