The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill's behavior is consistent with its stated purpose of auditing designs. The following security-positive patterns were observed:
[CONTROLLED_CAPABILITIES]: Destructive or write-heavy operations, such as perform_editing_operations in Figma or create_design_system_rules for repositories, are protected by mandatory user confirmation steps (Step 4 and Step 5.5).
[CREDENTIAL_HANDLING]: The instructions explicitly forbid the agent from repeating or storing user-provided credentials used to access authenticated URLs (Step 1
URL Input Spec).
[TRUSTED_SOURCES]: External data fetching via web_fetch for GitHub source code targets well-known and trusted domains (e.g., raw.githubusercontent.com).
[TRANSPARENCY]: The skill maintains clear communication with the user regarding inferred settings, detected frameworks, and confidence levels based on input types.

design-auditor