The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses the agent-browser tool to fetch external resources, including SVG logos, font files, and hero images from target websites to populate the generated kit.
[COMMAND_EXECUTION]: Utilizes platform-specific CLI tools od brand preview and od brand finalize to manage the rendering, validation, and registration of extracted brand kits.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data (DOM, CSS, and copy) from arbitrary websites.
Ingestion points: Harvesting content from live websites via agent-browser snapshots as described in SKILL.md.
Boundary markers: The instructions include a safety reminder to treat page content as untrusted evidence, though technical delimiters for the scraped content are not specified.
Capability inventory: The agent has file_write permissions to save assets and can execute od brand commands.
Sanitization: The included brand.html template implements an esc() function to escape HTML entities, mitigating XSS risks when rendering extracted data.

brand-extract