The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/agent.py and documentation in references/api-reference.md list dependencies on the requests and stix2 Python packages required for network communication and data formatting.
[DATA_EXFILTRATION]: The skill extracts security artifacts (IPs, domains, hashes) from user-provided local files and transmits them to external reputation services for enrichment. The services contacted include VirusTotal (virustotal.com), MalwareBazaar (abuse.ch), and AbuseIPDB (abuseipdb.com). This behavior is documented and consistent with the skill's incident response use case.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and processes raw, potentially untrusted incident data.
Ingestion points: Raw data is ingested through the input-file and input-text arguments in scripts/agent.py.
Boundary markers: No specific delimiters or instructions to ignore instructions within the ingested text are present.
Capability inventory: The script performs file reads, network API requests, and file writes.
Sanitization: The skill uses validation via specific regular expressions for IPs, domains, hashes, and URLs to filter input, which significantly limits the scope of processed content but does not provide complete sanitization against all forms of adversarial data.

collecting-indicators-of-compromise