The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents the idapython_snippet and idapython_file functions, which facilitate the execution of arbitrary Python code through SQL queries. This allows for runtime script execution within the IDA environment.
[COMMAND_EXECUTION]: The documentation includes functions that interact directly with the host file system, such as load_file_bytes for reading file contents into database memory, and gen_listing or gen_cfg_dot_file for writing data to host paths. These capabilities allow the agent to read from and write to arbitrary locations on the machine.
[INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted data sources including disassembly, decompilation output, and binary metadata (names, comments). Maliciously crafted content within these sources could potentially influence the agent's behavior. Ingestion points: decompile(), comments table, names table, and binary bytes data. Boundary markers: None identified in the provided documentation to delimit untrusted data from instructions. Capability inventory: idapython_snippet (Python execution), gen_listing (file system write), load_file_bytes (file system read), and the Bash tool. Sanitization: No explicit sanitization or validation logic is described for handling external binary data before it is presented to the agent context.

functions