The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/agent.py contains hardcoded default credentials ('admin' for both username and password) in the connect_gvm function signature and as default values for command-line arguments, which promotes the use of insecure, predictable authentication.
[EXTERNAL_DOWNLOADS]: The skill requires external Python libraries python-gvm and lxml, as documented in SKILL.md and references/api-reference.md, which are fetched from public package registries during setup.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it processes untrusted vulnerability data from external reports without sufficient security controls.
Ingestion points: External data enters the agent context in scripts/agent.py through load_data (JSON) and gmp.get_report (XML).
Boundary markers: No delimiters or explicit instructions are provided to the agent to disregard commands potentially embedded in the processed vulnerability reports.
Capability inventory: The script scripts/agent.py is capable of performing file system writes (Path.write_text) and initiating network connections (TLSConnection).
Sanitization: Vulnerability descriptions and names are truncated to 200 characters but are not escaped, filtered, or validated for malicious instructions.
[DATA_EXFILTRATION]: The skill facilitates network connections to user-specified hosts using the TLSConnection class in scripts/agent.py. This functionality allows the script to communicate with external, non-whitelisted domains.

implementing-vulnerability-management-with-greenbone