The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/agent.py contains hardcoded default credentials (admin/admin) within the connect_gvm function and as default values for the command-line argument parser. While these are common defaults for the service, their presence in the script poses a risk if not updated by the user.
[METADATA_POISONING]: There is a discrepancy between the author name mahipal in the SKILL.md manifest and the copyright holder mukul975 in the LICENSE file. Furthermore, the LICENSE file cites a future copyright year of 2026, which is inconsistent with current metadata standards.
[INDIRECT_PROMPT_INJECTION]: The skill processes vulnerability scan reports from external sources, presenting an attack surface where malicious input could influence agent behavior.
Ingestion points: The scripts/agent.py script reads external data via the load_data function for JSON files and the parse_report_xml function for XML reports.
Boundary markers: Absent; the script does not implement specific delimiters or instructions to isolate the parsed vulnerability data from the agent's operational logic.
Capability inventory: The scripts/agent.py script includes file-writing capabilities through Path.write_text for report generation.
Sanitization: The script truncates vulnerability descriptions to 200 characters but does not perform validation or escaping of the ingested content.
[EXTERNAL_DOWNLOADS]: The skill references and installs the python-gvm library and links to official Greenbone documentation. These resources are associated with a well-known service and are documented neutrally.

implementing-vulnerability-management-with-greenbone