The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by treating external documentation as "non-negotiable ground truth."
Ingestion points: Reads DESIGN.md and PRODUCT.md from the project root directory.
Boundary markers: None identified; instructions command the agent to read the files "fully" and follow "every token."
Capability inventory: The skill has the ability to write HTML files to the local filesystem (in mocks/ and accepted/ directories) and execute browser automation commands.
Sanitization: The skill lacks explicit sanitization or instructions to ignore potential directives embedded within the grounding files.
[COMMAND_EXECUTION]: The skill performs shell-based operations to automate design reviews.
Command logic: Executes agent-browser for resizing, screenshotting, and capturing snapshots of generated HTML files.
System Interaction: Uses shell redirection (>) to save browser snapshots to disk.

mock