11-csoc-automation
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted security data which could contain malicious instructions.
- Ingestion points: Data enters the environment via SIEM alert exports and EDR logs (e.g.,
alerts.json,siem_export.csv) which are parsed by thescripts/alert_triager.pyscript. - Boundary markers: The
SKILL.mdfile includes an 'Evidence contract' directing the agent to treat inputs as untrusted artifacts, although the Python scripts do not implement technical delimiters or isolation for the ingested text. - Capability inventory: The skill includes scripts that perform file system read and write operations to process alerts and generate Markdown reports.
- Sanitization: The triage script performs keyword matching on alert titles and descriptions without prior sanitization or filtering of the content.
Audit Metadata