The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill researches brand identities by retrieving information from external sources (web searches), which creates a surface for indirect prompt injection.
Ingestion points: Research findings from web search tools as described in the 'Research' step of SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious embedded instructions in the retrieved research data.
Capability inventory: The skill writes files to the working directory (brand-brief.md, identity.js, brand-preview.pdf) and executes a browser-based rendering process via Playwright in engine/render.js.
Sanitization: No explicit sanitization or filtering of retrieved web content is performed before the data is processed into the brand brief.
[EXTERNAL_DOWNLOADS]: The rendering engine fetches font assets from the Google Fonts CDN.
The engine/index.js file generates HTML containing links to 'fonts.googleapis.com'. This is a well-known service used for its intended purpose of providing typography specimens for the brand preview.
[COMMAND_EXECUTION]: The skill uses Playwright to launch a headless browser instance for PDF generation.
The engine/render.js script utilizes the 'playwright' library to launch a Chromium instance (chromium.launch). This is a standard and expected implementation for the skill's primary purpose of generating visual previews.

brand-system