The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/agent.py script downloads the official MITRE ATT&CK Enterprise dataset from https://raw.githubusercontent.com/mitre/cti/master/enterprise-attack/enterprise-attack.json. This is a well-known, trusted industry source for cybersecurity threat intelligence and is used purely for data lookup and technique mapping.
[COMMAND_EXECUTION]: The documentation describes various offensive security commands (e.g., using Cobalt Strike, RDP, or Kerberoasting). However, these are provided as instructional guidance for human operators and are not executed by the skill's scripts. The Python script only performs data processing and reporting.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external JSON data (the MITRE ATT&CK dataset) and processes it to generate reports. While this is an ingestion point for untrusted data, the script uses standard JSON parsing and focuses on extracting structured fields (IDs, names, tactics) rather than executing instructions from the data. The risk of this data influencing the agent's behavior is low due to the structured nature of the processing.

executing-red-team-exercise