skills/aws-samples/sample-agent-skills-for-builders/agentcore-browser-web-scraping/Gen Agent Trust Hub
agentcore-browser-web-scraping
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill uses a set of fixed code primitives (tools) that the LLM can parameterize but not modify. This architecture prevents the model from executing arbitrary JavaScript or scripts within the cloud browser environment.
- [SAFE]: Implements a robust SSRF (Server-Side Request Forgery) guard within the navigation tool. It enforces HTTPS, rejects IP literals, and prevents access to private IP ranges or AWS Instance Metadata Service (IMDS) endpoints by validating resolved IPs before every navigation call.
- [SAFE]: Securely manages sensitive credentials by leveraging Amazon Bedrock AgentCore Browser Profiles for persistent login states and AWS Secrets Manager ARNs for proxy authentication. This ensures that raw credentials and session tokens never enter the LLM's prompt or the application logs.
- [SAFE]: Data extraction logic in
extract_by_selectoris implemented using a constant JavaScript template where model-supplied selectors are passed as data arguments, avoiding unsafe string concatenation oreval()calls. - [SAFE]: External resources and installation paths point to official AWS repositories and documentation, aligning with the stated author context.
Audit Metadata