building-attack-pattern-library-from-cti-reports

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill processes external Cyber Threat Intelligence (CTI) reports containing untrusted data, which presents a surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the agent context through CTI report text files read via the --report-files CLI argument in scripts/agent.py and the parse_report method in SKILL.md.
  • Boundary markers: No delimiters or 'ignore' instructions are used to distinguish report content from agent instructions during processing.
  • Capability inventory: The skill's capabilities include reading local files and writing JSON-formatted library data and detection templates to the local filesystem.
  • Sanitization: Extracted text snippets (such as the 'source_sentence' field) are included in the generated output without sanitization or escaping.
  • [EXTERNAL_DOWNLOADS]: The skill fetches public MITRE ATT&CK enterprise data from the official MITRE GitHub repository to facilitate technique mapping. This is a reference to a well-known and trusted service.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 02:48 AM