editorial-card-screenshot
Warn
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a shell script (
scripts/capture_card.sh) to execute browser binaries (google-chrome,chromium, or a path defined by theCHROME_BINenvironment variable) in headless mode to take screenshots of local HTML files. It also references a secondary script (scripts/trim_card_bottom.sh) for post-processing rendering results.\n- [EXTERNAL_DOWNLOADS]: The skill references Google Fonts (fonts.googleapis.com), which is a well-known and trusted service. This resource is fetched when the headless browser renders the HTML card.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied information and embeds it into HTML info cards, which are then rendered by a headless browser. This creates an attack surface for instructions hidden in user data.\n - Ingestion points: User-provided text, headlines, and content units used to generate the HTML cards (
SKILL.md).\n - Boundary markers: The instructions do not define delimiters or specific warnings to prevent the agent from interpreting embedded instructions within the user data.\n
- Capability inventory: The skill can execute shell scripts and run a headless browser (
scripts/capture_card.sh,SKILL.md).\n - Sanitization: There is no logic or instruction for sanitizing user-provided content before it is inserted into the HTML templates, which could allow the execution of JavaScript in the browser environment if the input contains malicious tags.\n- [DYNAMIC_EXECUTION]: The workflow involves generating a local HTML file from templates and then invoking an external process (a web browser) to interpret and render that content. While not traditional code execution, this pattern executes the logic and layout described in the generated HTML/CSS/JS.
Audit Metadata