The Agent Skills Directory

[SAFE]: The skill instructions define a geopolitical analysis workflow using trusted news organizations such as Reuters, Bloomberg, and Cailian Press for data retrieval.\n- [SAFE]: The included script scripts/format_report.py is a formatting utility that uses only the Python standard library and does not perform any dangerous system calls, network operations, or file system modifications.\n- [SAFE]: Data persistence is limited to logging analysis summaries in a local memory/ directory to provide historical context between sessions.\n- [SAFE]: Indirect Prompt Injection Surface Assessment: 1) Ingestion points: Fetches geopolitical news content via external web searches (SKILL.md). 2) Boundary markers: Utilizes structured Markdown templates and predefined report formats (report_template.md). 3) Capability inventory: Web search and local file read/write access restricted to the skill's own memory folder. 4) Sanitization: No explicit sanitization of news content is described, but the agent's capabilities are restricted to generating financial reports.

geopolitical-driver