xiaohongshu-images
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (
scripts/screenshot.py) to automate browser-based screenshots. The agent is instructed to sanitize input-derived strings (like article titles) before using them in shell commands, reducing injection risks. - [EXTERNAL_DOWNLOADS]: The skill's CSS templates reference Google Fonts (
fonts.googleapis.com), which is an industry-standard, well-known service for typography. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data from user inputs and URLs.
- Ingestion points: User-provided text, markdown files, and content fetched from external URLs.
- Boundary markers: Prompt templates in the
prompts/directory utilize<user_content>tags to delimit untrusted data. - Capability inventory: Local file system writes to
~/Dev/obsidian/, execution of thescreenshot.pyscript via Playwright, and invocation of the/baoyu-cover-imageskill. - Sanitization: The skill relies on the agent to parse and reformat content into a predefined HTML structure, rather than rendering raw user input directly, and includes explicit instructions for sanitizing file paths.
Audit Metadata