The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data such as threat feeds, indicators of compromise (IOCs), and malware metadata. While this represents a potential attack surface, the current implementation is safe.
Ingestion points: The ingest method in the IOCEnrichmentPipeline class within SKILL.md accepts a list of raw strings for processing.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined in the prompt instructions for the agent when processing this data.
Capability inventory: The provided Python script does not utilize network operations, file system access, or subprocess execution; it is limited to string parsing and structured data generation.
Sanitization: The script implements input validation using regular expressions and the standard ipaddress library to filter and classify data before enrichment.
[COMMAND_EXECUTION]: The skill provides Sigma rules that describe command-line patterns (e.g., PowerShell download cradles) for detection purposes. These are static data definitions and do not involve execution by the agent or the skill's own environment.
[OBFUSCATION]: The YARA and Sigma rules contain hex patterns, base64 strings, and detection logic for obfuscated commands. These are legitimate components of threat detection signatures and do not represent obfuscation of the skill's own logic or instructions.

Threat Intelligence Analyst