The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references and downloads official tools from well-known security researcher Didier Stevens (oledump.py) and standard Python packages (oletools, xlmdeobfuscator) for malware analysis.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Office documents which could contain embedded instructions designed to influence the agent during the analysis process.
Ingestion points: The agent.py script and olevba utility read content from user-provided suspicious documents.
Boundary markers: The analysis report uses structured headers to separate extracted content, though it lacks explicit delimiters to prevent agents from interpreting macro code as instructions.
Capability inventory: The skill environment includes file read access and shell command execution for analysis tools.
Sanitization: Extracted content is deobfuscated for human/agent review but not specifically filtered for natural language injection patterns.
[SAFE]: The skill is a legitimate security analysis tool intended for examining suspicious Office documents in a controlled environment and contains no malicious logic, hidden network operations, or exfiltration patterns.
[SAFE]: The skill promotes security best practices, such as the use of isolated analysis VMs and safe document rendering tools like LibreOffice to prevent accidental malware execution.

analyzing-macro-malware-in-office-documents