The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the download and execution of a remote PowerShell installation script (install-atomicredteam.ps1) using IEX (Invoke-Expression). Automated scanners have flagged this specific URL (https://raw.githubusercontent.com/redcanaryco/invoke-atomicredteam/master/install-atomicredteam.ps1) as malicious (Botnet). This pattern allows for the silent execution of arbitrary code on the user's system.\n- [COMMAND_EXECUTION]: The skill's primary function is the execution of arbitrary attack techniques via Invoke-AtomicTest. These techniques involve high-risk actions such as credential dumping from LSASS memory, modification of system registry keys for persistence, and creation of scheduled tasks, which can compromise the integrity and security of the host.\n- [EXTERNAL_DOWNLOADS]: The skill downloads software and test libraries from external GitHub repositories (redcanaryco/invoke-atomicredteam). These external resources are not vetted and include scripts flagged as malicious, posing a significant supply chain risk.\n- [PROMPT_INJECTION]: The Python agent and associated workflows are vulnerable to indirect prompt injection due to their processing of external, potentially untrusted data files.\n
Ingestion points: agent.py reads YAML technique definitions and JSON detection results from user-defined paths.\n
Boundary markers: Absent; there are no delimiters to isolate the processed data from the agent's logic.\n
Capability inventory: The skill can execute shell commands via Invoke-AtomicTest, write files, and generate new executable PowerShell scripts based on external data.\n
Sanitization: There is no validation or sanitization of the data ingested from YAML or JSON files before it is processed or used to generate retest scripts.

performing-purple-team-atomic-testing