The Agent Skills Directory

[SAFE]: The skill provides structured instructions for auditing front-end code across five dimensions: accessibility, performance, theming, responsive design, and anti-patterns. It does not perform any sensitive operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted code (components and files) to perform audits. However, the risk is negligible because the skill lacks high-risk capabilities such as network exfiltration, arbitrary command execution, or file system modification; it only produces a text-based report for human review. No specific sanitization or boundary markers are defined for the input data, but the lack of dangerous tools mitigates the threat.
[EXTERNAL_DOWNLOADS]: The skill metadata references a public GitHub repository belonging to a well-known developer as a source for its guidelines. This is a legitimate reference for documentation and does not involve any automated script execution or package installation.
[COMMAND_EXECUTION]: The skill recommends using other agent commands (e.g., /normalize, /optimize, /harden) to address issues found during the audit. These are standard task-chaining instructions within the agent environment and do not involve unauthorized shell access or privilege escalation.

audit