The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a feature to scrape and analyze up to 100 tweets from X/Twitter profiles (Scenario E). This process introduces a surface for Indirect Prompt Injection, where malicious instructions embedded in the scraped tweets could attempt to override the agent's behavior during the analysis or report generation phase.
Ingestion points: Data is ingested from external Twitter profiles via scraping in SKILL.md (Scenario E).
Boundary markers: The instructions do not specify any delimiters or safety constraints to prevent the agent from following instructions found within the scraped content.
Capability inventory: The agent has the ability to write to the local filesystem (user-data/), navigate the web using computer-use, and generate HTML reports.
Sanitization: There is no evidence of sanitization or filtering applied to the external text before it is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on automated browser tools and the javascript_tool to interact with external web pages and extract data. In Scenario E, it specifically uses computer-use to navigate and screenshot pages, and claude-in-chrome to extract DOM elements and execute scripts. While these are platform-provided tools, their application on untrusted third-party websites represents a controlled but relevant execution risk.

x-mastery-mentor