The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the webcmd CLI via Bash to perform all browser interactions, including state inspection, navigation, and element manipulation.
[EXTERNAL_DOWNLOADS]: The skill instructions include a fallback gate that uses webcmd plugin search to find and offer the installation of external plugins, which involves downloading third-party code.
[DATA_EXFILTRATION]: The browser upload command allows the agent to attach local files to a website's file input. If the agent is compromised via injection, this could be used to exfiltrate sensitive local documents. Furthermore, the browser network command captures API traffic which may contain PII or session data.
[DYNAMIC_EXECUTION]: The browser eval command allows for the execution of arbitrary JavaScript within the browser's context. Although the documentation encourages read-only IIFEs, the command itself provides a mechanism for dynamic code execution.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from arbitrary websites through commands like state, extract, and fetch-browser. This creates a surface for indirect prompt injection where a malicious website could attempt to influence the agent's behavior.
Ingestion points: Untrusted data enters the agent context via browser state, browser extract, fetch-browser, and browser network (as described in SKILL.md).
Boundary markers: The instructions do not specify explicit delimiters or specific warnings to ignore instructions embedded in the web content.
Capability inventory: The agent has the ability to write to the local filesystem (Write, Edit tools), execute shell commands (Bash(webcmd:*)), and upload local files to external domains.
Sanitization: No specific sanitization, escaping, or filtering of instructions found within external web content is described.

webcmd-browser