The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes STIX/TAXII threat intelligence data from external servers, creating a surface for indirect prompt injection where malicious instructions could be embedded in threat indicators.
Ingestion points: scripts/agent.py and scripts/process.py utilize the taxii2client library to fetch STIX objects from remote TAXII collections.
Capability inventory: The scripts possess file-write capabilities, allowing them to save extracted indicators and threat reports to local JSON or CSV files via the --output and --csv-output arguments.
Boundary markers: The implementation lacks explicit boundary markers or directives to ignore instructions when parsing data retrieved from external feeds.
Sanitization: While scripts/process.py uses regular expressions to extract specific observable values (IPs, domains, hashes), other fields like object names and descriptions are processed and displayed without sanitization.
[DATA_EXFILTRATION]: The integration scripts perform network operations to retrieve threat intelligence data from remote endpoints.
The scripts/agent.py and scripts/process.py tools facilitate outbound HTTP connections to TAXII servers using the taxii2-client and requests libraries. Although the documentation focuses on well-known services such as MITRE ATT&CK, the tools allow connection to any user-specified URL.

implementing-stix-taxii-feed-integration