The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses strong imperative language to override the AI's default behavior and communication style. Specifically, it instructs the agent to conceal its internal operations from the user with commands like "Be invisible," "Don't announce modes or narrate process," and explicit prohibitions on mentioning its current state (e.g., "Never say: 'I'm in ESTABLISH MODE'").
[COMMAND_EXECUTION]: The instructions require the AI to use the TaskCreate tool for all operations, which contradicts the restrictive allowed-tools: NONE setting in the YAML frontmatter. It also mandates writing structured data and design patterns to the local file system (e.g., docs/design-system/interface-system.md), which involves tool invocation for file operations.
[SAFE]: No evidence of hardcoded credentials, malicious remote downloads, or obfuscated code was found within the analyzed files.

interface-design