opencli-browser

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to interact with and extract data from arbitrary web pages, which introduces a surface for indirect prompt injection attacks where malicious instructions embedded in a web page could influence the agent's behavior.
  • Ingestion points: Untrusted data enters the agent's context through several commands including browser state, browser get text, browser extract, and browser network (SKILL.md).
  • Boundary markers: The instructions do not specify the use of delimiters or provide explicit warnings to the agent to ignore or isolate instructions found within the processed web content.
  • Capability inventory: The skill utilizes the Bash(opencli:*), Read, Edit, and Write tools, granting the agent significant capabilities to execute commands and interact with the file system based on its interpretation of page content.
  • Sanitization: While the opencli tool returns structured JSON envelopes, the skill does not implement explicit sanitization or filtering of the natural language data extracted from the DOM before it is presented to the agent.
  • [DYNAMIC_EXECUTION]: The browser eval <js> command allows for the execution of arbitrary JavaScript within the context of the active browser tab. Although the documentation advises the agent that this command should be read-only, the underlying capability exists to dynamically execute code that could modify page state or interact with site internals.
  • [DATA_EXFILTRATION]: The skill includes a browser upload [target] <file...> command which allows the agent to attach local files to browser inputs. This capability could be leveraged to exfiltrate sensitive local data if an agent is misdirected to upload specific files (e.g., credentials or configuration) to an external site.
  • [EXTERNAL_DOWNLOADS]: The documentation provides a reference to the official Google Chrome Web Store to install the necessary browser extension required for the skill to function. This reference targets a well-known service.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 03:28 AM