daily-news-caster

Warn

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install a dependency from an unverified 3rd-party GitHub repository (https://github.com/cclank/news-aggregator-skill). While the tts skill is sourced from the vendor's own repository, the aggregator source is untrusted.
  • [REMOTE_CODE_EXECUTION]: The skill relies on downloading and executing external scripts (fetch_news.py) from an untrusted repository. This pattern allows for arbitrary code execution on the user's machine if the remote repository content is modified or malicious.
  • [COMMAND_EXECUTION]: The skill uses npx, python3, and bash to execute local and downloaded scripts. Specifically, it executes fetch_news.py and tts.sh, which are external dependencies.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It ingests data from external news sources (e.g., Hacker News, GitHub) via fetch_news.py (File: SKILL.md Step 2) and interpolates this content directly into a podcast script (File: SKILL.md Step 3) without sanitization, boundary markers, or instructions to ignore embedded commands. This data is then processed by the LLM to generate the script, allowing malicious news content to influence agent behavior. The skill maintains significant capabilities, including subprocess execution (python3, bash) and file system access.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 7, 2026, 08:16 AM