The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the oletools and xlmdeobfuscator packages from the Python Package Index. It also directs users to download oledump.py from the official blog of a well-known security researcher. These dependencies are standard requirements for the document analysis tasks described.
[COMMAND_EXECUTION]: The provided workflows and the agent.py script facilitate the execution of several command-line analysis tools, including olevba, oleid, and oledump. These tools are used for the static inspection of OLE streams and document structures to identify suspicious indicators without executing the document's payload.
[PROMPT_INJECTION]: The skill's primary function is to process untrusted data from potentially malicious Office documents. This exposes the agent to indirect prompt injection, where strings or code extracted from a document might be interpreted as commands by the AI agent during the analysis report generation.
Ingestion points: Untrusted data enters the environment through sys.argv and is processed by VBA_Parser and zipfile in agent.py.
Boundary markers: There are no explicit delimiters or warning instructions used to isolate the extracted macro strings from the agent's internal logic.
Capability inventory: The agent possesses file system access and the ability to execute subprocesses as part of the analysis workflow.
Sanitization: While the script performs regex-based deobfuscation to aid human analysis, it does not sanitize the extracted content to prevent it from being interpreted as instructions by the language model.

analyzing-macro-malware-in-office-documents