The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from the local repository for design alignment.
Ingestion points: The skill searches for and reads content from files matching 'design system', 'ui guide', or 'style guide' within the workspace as described in the 'Plan' section of SKILL.md.
Boundary markers: No specific delimiters or 'ignore instructions' warnings are specified for the ingested content.
Capability inventory: The skill has the capability to modify and delete files during the normalization and cleanup phases as described in the 'Execute' and 'Clean Up' sections of SKILL.md.
Sanitization: No sanitization or filtering of the content retrieved from the search is described.
Context: This ingestion is a core component of the skill's primary UI normalization purpose, and it includes a task-level safeguard to ask for clarification if principles are unclear.
[SAFE]: The skill operates within the expected scope of a frontend design and refactoring tool. It does not contain obfuscated code, remote execution patterns, or attempts to exfiltrate sensitive data. All instructions are task-specific and aim to ensure consistency with existing project standards.

normalize