The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the bb-browser binary and utilizes the Bash(bb-browser:*) tool to execute arbitrary commands within that tool's namespace. This allows the agent to perform web scraping and system-level operations defined by the binary's capabilities.
[EXTERNAL_DOWNLOADS]: The instruction set includes the command bb-browser site update, which is described as a way to 'pull community adapters'. These adapters are external logic or configuration files downloaded from a remote source. If these community-contributed files are not subject to a rigorous verification process, they could contain malicious logic that exploits the browser environment or the agent's permissions.
[PROMPT_INJECTION]: The primary function of the skill is to ingest data from a vast array of external websites (social media, news, developer forums, etc.). This content is untrusted and provides a significant surface for indirect prompt injection attacks.
Ingestion points: Data is entered into the agent's context via the output of bb-browser site commands across various platforms (SKILL.md).
Boundary markers: The instructions do not define any delimiters or system messages to help the agent distinguish between data and potential instructions embedded in the scraped content.
Capability inventory: The agent has the ability to execute Bash commands via the bb-browser tool and process the resulting data in its reasoning loop.
Sanitization: There is no evidence of content sanitization, filtering, or validation before the scraped data is processed by the AI.

bb-browser-openclaw