The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository during its 'Auto' mode, scanning logos, stylesheets, and documentation files to infer brand values. To mitigate risks, the skill uses a strict provenance system, labeling all information as answered, context, researched, or invented. All extracted information requires user confirmation and a 'lock' before being committed to the final specification, providing a human-in-the-loop security boundary.
[DYNAMIC_EXECUTION]: The skill generates code in the form of a static website consisting of HTML, CSS, and a small vanilla JavaScript file for navigation and interaction. The instructions mandate that the site must remain fully readable even if JavaScript is disabled. It also uses the standard zip utility to package a bootstrap kit for the user. These are standard development tasks and do not involve executing untrusted or remote code.
[COMMAND_EXECUTION]: The skill uses the system's zip command-line tool to create a 'bootstrap kit' (a compressed archive of the generated tokens and specifications). This is a standard utility usage for the intended purpose of the skill.

brand-guidelines