milan-news-briefing
Warn
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the subprocess module for system management and automation. Specifically, scripts/manage_cron.py interacts with the crontab utility to establish and maintain scheduled execution of the briefing tasks, which constitutes a persistence mechanism. Other scripts like daily_ops.py and run_briefing.py use subprocess.run to launch the news generation pipeline as child processes.\n- [EXTERNAL_DOWNLOADS]: The skill downloads data from a variety of external sources. The src/news_briefing/pipeline.py and fetch.py modules retrieve information from the Open-Meteo API, numerous RSS news feeds (e.g., ANSA, BBC, MIT Technology Review), and perform web searches via Brave Search and DuckDuckGo. Health checks in scripts/check_feeds.py also perform network requests to verify these sources.\n- [DATA_EXFILTRATION]: scripts/daily_ops.py contains a function to transmit execution logs and status summaries to an external webhook URL provided at runtime. This can include task completion status, attempt history, and log snippets (the last 1500 characters of the process output) for remote monitoring.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its ingestion of untrusted external content.\n
- Ingestion points: External data from RSS feeds and web search results are retrieved in src/news_briefing/fetch.py and processed in pipeline.py.\n
- Boundary markers: No explicit delimiters or instructions are used to separate retrieved news content from the agent's internal instructions in the generated report.\n
- Capability inventory: The skill possesses capabilities to execute commands via subprocess and transmit data via webhooks.\n
- Sanitization: No sanitization or filtering of instructions within news titles or descriptions was observed in parse.py or pipeline.py.
Audit Metadata