geopolitical-risk

Pass

Audited by Gen Agent Trust Hub on Nov 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external news sources and the ACLED API, creating a surface for indirect prompt injection where malicious content from these sources could influence the agent.
  • Ingestion points: analyze_geopolitical_news (web content) and fetch_acled_events (API data) in SKILL.md.
  • Boundary markers: Absent.
  • Capability inventory: Network read access for data fetching.
  • Sanitization: No explicit sanitization or boundary markers are defined for the content passed to the agent for analysis.
  • [DYNAMIC_EXECUTION]: The query_gdelt_events function in SKILL.md constructs a SQL query for BigQuery using f-strings to interpolate variables such as country_code, event_type, and dates. This pattern is vulnerable to SQL injection if the agent populates these arguments from untrusted user queries without proper parameterization.
  • [EXTERNAL_DOWNLOADS]: The skill fetches the official GPR Index data from a recognized academic resource (matteoiacoviello.com) and conflict data from the ACLED API to facilitate geopolitical analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 21, 2026, 05:36 PM