cmd-rss-feed-generator
Warn
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill's primary workflow involves the agent generating executable Python scripts (e.g., in
feed_generators/) which are then intended to be run by the user or an automated process. This script generation and subsequent execution of non-static code is a form of dynamic code execution. - [COMMAND_EXECUTION]: The instructions include several shell commands for managing and running the scrapers, such as
python feed_generators/new_site_blog.pyandmake feeds_new_sitevia a Makefile. - [EXTERNAL_DOWNLOADS]: The generated scrapers are designed to fetch content from arbitrary external blog URLs, which introduces risk depending on the target domain and the processing of the retrieved content.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted HTML data from external websites. (1) Ingestion points:
fetch_content(url)and local HTML file imports. (2) Boundary markers: None specified to delimit or sanitize scraped content. (3) Capability inventory: Python execution, file system writes (feeds/*.xml), and network access for scraping. (4) Sanitization: No explicit sanitization or filtering of the scraped content is mentioned beyond basic HTML parsing.
Audit Metadata