The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system-level commands through subprocess.run to interact with package managers like apt, yum, and dnf, as well as Windows PowerShell. These commands are used to check for pending updates and are essential for the skill's patch management functionality. Additionally, the workflow documentation includes a curl command using the -k or --insecure flag when communicating with a Nessus server, which bypasses SSL certificate validation and could allow for data interception in transit. \n- [EXTERNAL_DOWNLOADS]: The agent retrieves the Known Exploited Vulnerabilities (KEV) catalog from CISA's official government website. This retrieval is part of the skill's intended security auditing logic to cross-reference and prioritize vulnerabilities. \n- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data from vulnerability scan results and asset inventories. \n
Ingestion points: Data enters the system via CSV files loaded in scripts/process.py using pandas.read_csv. \n
Boundary markers: There are no explicit delimiters or instructions to the agent to ignore any prompts that might be embedded within the CSV fields. \n
Capability inventory: The skill includes scripts capable of system command execution (agent.py) and file generation (process.py). \n
Sanitization: While the skill uses standard parsing libraries like pandas and json to handle the data, it does not implement specific filtering or sanitization to prevent malicious instructions hidden in the data from being interpreted.

implementing-patch-management-workflow