The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute a local Python script (scripts/gate.py) to perform automated browser-based UI testing and design verification.
[EXTERNAL_DOWNLOADS]: The skill depends on the playwright library and requires the installation of the Chromium browser engine via playwright install. These are official tools from a well-known technology provider.
[DYNAMIC_EXECUTION]: The gate.py script executes a significant block of JavaScript logic within the rendered page context to calculate layout properties, contrast ratios, and detect specific visual patterns.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content (HTML files or URLs) and extracts text for reporting purposes, creating an attack surface for indirect instruction injection.
Ingestion points: The target URL or local HTML file path provided to the gate.py script as documented in SKILL.md.
Boundary markers: Absent. Extracted text from elements (e.g., inner text and alt attributes) is interpolated directly into finding reports without delimiters or warnings to the agent.
Capability inventory: The skill has the capability to run scripts, launch browsers, and generate files in the .tastegate output directory.
Sanitization: The script truncates extracted text to 48 characters, which provides a minor limitation but does not prevent the inclusion of malicious instructions in the reported findings.

tastegate