The Agent Skills Directory

Local Workflow Automation: The skill utilizes subprocess calls to interact with the system and other developer tools. Specifically, it manages local network ports using utilities like lsof and interacts with the claude CLI to perform triggering evaluations. These operations are restricted to the local environment and support the skill's core functionality.
Dynamic Resource Generation: To evaluate skill triggering, the tool dynamically generates temporary skill configurations in the .claude/commands/ directory. These resources are used strictly for testing during the optimization loop and are cleaned up after execution.
Local Visualization Server: A built-in script serves a local HTTP dashboard on 127.0.0.1 to provide a visual interface for reviewing evaluation results. This server is designed for local use to ensure user data remains within the local environment.
External Dependency Usage: The skill depends on the anthropic Python SDK and the claude CLI. These are official vendor resources used for generating improved skill descriptions and testing triggering behavior.
Data Processing Surface: The skill processes user-provided evaluation queries and feedback. While this represents a potential surface for indirect prompt injection, the instructions include guidance for human review and automated grading to verify output quality.

skill-creator