The Agent Skills Directory

[SAFE]: No malicious patterns, hardcoded credentials, or direct prompt injections were detected in the skill documentation.
[NO_CODE]: The skill consists of documentation and interface definitions for existing MCP tools; no additional scripts or executable code are included in the package.
[PROMPT_INJECTION]: The skill defines an interface that reads untrusted data from the web, creating a potential surface for indirect prompt injection. 1. Ingestion points: The read_page, read_console_messages, and read_network_requests tools ingest content from external websites into the agent's context. 2. Boundary markers: No delimiters or explicit instructions to ignore instructions embedded in ingested content are present. 3. Capability inventory: The skill possesses high-privilege capabilities including form_input, computer (mouse/keyboard simulation), and javascript_tool (JS execution in-browser). 4. Sanitization: No sanitization or validation of ingested web content is described in the documentation.

claude-browser