detection-engineering

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves detection rules, testing samples, and security utilities from recognized public repositories including SigmaHQ, Yara-Rules, and projects maintained by well-known security researchers (e.g., Florian Roth, S. Bousseaden). It also downloads pre-built security scanning tools from Yamato Security. These resources are standard components of the detection engineering ecosystem.\n- [COMMAND_EXECUTION]: The instructions guide the agent in executing various command-line security tools such as yara, sigma-cli, hayabusa, loki, and yarGen. These tools are utilized for their primary intended purpose: scanning log files, identifying malware signatures, and managing detection-as-code workflows.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could potentially contain malicious instructions targeting the agent's logic.\n
  • Ingestion points: The skill ingests external data including threat intelligence reports, Windows Event Logs (EVTX), and community-submitted Sigma and YARA rules as described in SKILL.md and the provided guides.\n
  • Boundary markers: Employs formal validation steps such as YAML schema parsing (sigma-cli parse) and multi-stage testing in staging environments to delimit rule content from execution logic.\n
  • Capability inventory: The execution environment permits tools like Bash, Write, Edit, and WebFetch, which represent a potential surface if embedded instructions are accidentally followed.\n
  • Sanitization: Relies on structural schema validation, peer review of detection logic, and positive/negative corpus testing to identify and filter out problematic or deceptive content within ingested rules.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 04:36 PM