The Agent Skills Directory

[SAFE]: The skill serves as an internal template for visual evidence capture and regression testing using the agent-browser tool. All documented workflows (screenshots, video recording, HAR capture) are standard for UI/UX testing and debugging.
[DATA_EXFILTRATION]: The skill facilitates the creation of HAR files (network traces) via agent-browser network har start. While HAR files can contain sensitive information like session cookies and authorization headers, the skill instructions are directed at local debugging and performance auditing. The reference to the external analyzer har.tech is an industry-standard practice for HAR visualization.
[PROMPT_INJECTION]: As the skill involves opening and interacting with external URLs (agent-browser open <url>), it is inherently susceptible to indirect prompt injection where a malicious website could attempt to influence the agent's behavior.
Ingestion points: Browser interaction with untrusted URLs defined in SKILL.md workflows.
Boundary markers: Absent; the instructions do not explicitly provide delimiters for website content or warnings to ignore embedded instructions.
Capability inventory: Screenshot capture, DOM snapshotting, video recording, and network tracing via the agent-browser toolset.
Sanitization: None; the agent processes the live DOM and network traffic of the target page.

agent-browser-visual