cloudflare-browser-rendering
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The
templates/ai-enhanced-scraper.tsfile provides an example of AI-enhanced scraping that is vulnerable to indirect prompt injection. It retrieves content from user-provided URLs and includes it directly in an LLM prompt. - Ingestion points: External HTML content is fetched from URLs passed as query parameters in
templates/ai-enhanced-scraper.ts. - Boundary markers: The prompt lacks distinct delimiters to separate the AI's instructions from the potentially untrusted HTML data.
- Capability inventory: The skill has the capability to interact with browser bindings, KV storage, and R2 buckets as configured in the project.
- Sanitization: There is no sanitization or validation of the fetched HTML before it is processed by the AI model.
- [EXTERNAL_DOWNLOADS]: The
scripts/check-versions.shscript makes requests to the npm registry to fetch version metadata for several packages. - [COMMAND_EXECUTION]: The
scripts/check-versions.shfile is a shell script intended for manual execution to verify dependency versions.
Audit Metadata