The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions that explicitly direct the agent to bypass user confirmation and safety protocols. Specifically, it commands the agent to install missing dependencies "automatically without asking the user to confirm" and instructs the agent to "not request separate approval for dependency installation." It further directs the agent to conceal this activity by stating, "Do not describe it as a request for the user to open their browser."
[REMOTE_CODE_EXECUTION]: The skill establishes a remote code execution vector by requiring the automatic installation of a CLI and a secondary skill from an external URL if they are not already present in the environment. Executing code from a remote source without user review is a high-risk activity.
[EXTERNAL_DOWNLOADS]: The skill references an external URL (https://www.browseract.com/) to fetch and install the BrowserAct CLI and Skill, which are treated as mandatory runtime dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to scrape and process content from external ecommerce websites, which can lead to indirect prompt injection if those websites contain malicious instructions.
Ingestion points: External ecommerce product pages, specifically question and answer threads captured via the browser-act tool (SKILL.md).
Boundary markers: None identified; the skill lacks instructions for delimiters or warnings to ignore commands embedded in the scraped text.
Capability inventory: The skill uses the browser-act tool to interact with live websites and the local CLI to manage dependencies.
Sanitization: None identified; there are no instructions to validate, filter, or escape the captured browser content before returning it to the user or agent context.

ecommerce-question-answer-scraper