The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes Windows Sysmon logs (EVTX or JSON format) which contain arbitrary, attacker-controlled strings such as command-line arguments and filenames.
Ingestion points: Untrusted log data is loaded into the agent's context through the load_events and parse_sysmon_evtx functions in scripts/agent.py.
Boundary markers: The skill does not implement boundary markers or instructions to the AI agent to ignore natural language instructions embedded within the log entries.
Capability inventory: The skill possesses file-reading capabilities and produces output containing raw command-line strings that are returned to the AI agent.
Sanitization: There is no evidence of sanitization or escaping of extracted strings before they are processed by the AI, which could allow an attacker to trigger unintended behavior by placing prompt injection payloads in command-line arguments.
[COMMAND_EXECUTION]: The documentation in SKILL.md provides several shell and PowerShell command snippets intended for system configuration (e.g., installing Sysmon and applying AppLocker policies). These commands require administrative privileges to execute and could be misused if executed without user review.

detecting-living-off-the-land-attacks