The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted and malicious samples including executables, documents, and scripts. 1. Ingestion points: Untrusted data enters the agent context through files loaded from paths like /samples/ and /quarantine/ into various analyzer classes like MalwareAnalyzer and PEAnalyzer (SKILL.md). 2. Boundary markers: The skill body does not define specific delimiters or instructions to treat analyzed content, such as extracted strings or deobfuscated code, as non-executable data, increasing the risk that the agent may interpret embedded content as instructions. 3. Capability inventory: The skill allows for extensive file reads, YARA scanning, and script deobfuscation across the local filesystem (SKILL.md). 4. Sanitization: There is no evidence of sanitization or filtering of extracted metadata, strings, or IOCs before they are presented in analysis results or reports.

malware-forensics