The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing content from arbitrary external URLs, which creates a significant surface for indirect prompt injection attacks.
Ingestion points: The skill uses browser_navigate and browser_screenshot in Phase 1 to capture data from a URL provided by the user.
Boundary markers: The instructions do not define clear delimiters or include warnings to the agent to disregard potential instructions embedded within the target website's text or metadata.
Capability inventory: The agent possesses capabilities to write files (creating BRAND-ANALYSIS.md, SITE.md, and React/Next.js components) and uses browser automation tools.
Sanitization: There is no evidence of sanitization, filtering, or safety checks performed on the visual or textual data retrieved from the website before it is used to influence further agent actions.
[EXTERNAL_DOWNLOADS]: The skill initiates connections to external, untrusted web environments through browser automation.
Evidence: Phase 1 instructions specify using Playwright MCP to browser_navigate to a user-supplied URL to capture visual references.

page-remake