agent-browser
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
agent-browserpackage from the NPM registry. - [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to executeagent-browserCLI commands for browser automation and system interaction. - [DYNAMIC_EXECUTION]: The skill uses a dynamic instruction loading mechanism (
agent-browser skills get core) to fetch workflows and command references at runtime, meaning the operational instructions are not fully contained within the static skill file. - [INDIRECT_PROMPT_INJECTION]: The skill enables browser-based tasks like web scraping and form interaction, which exposes the agent to potential indirect prompt injection from untrusted web content.
- Ingestion points: Interacting with external websites, accessibility-tree snapshots, and data extraction via the browser CLI.
- Boundary markers: The skill stub does not define explicit delimiters to separate trusted instructions from untrusted website content.
- Capability inventory: The skill has access to the
Bashtool, network connectivity (through the browser), and functionality for session management and authentication storage. - Sanitization: There are no mentioned procedures for sanitizing or escaping data retrieved from the web before it is processed by the agent.
Audit Metadata