The Agent Skills Directory

[SAFE]: The skill is a complex diagnostic framework designed to improve the quality of other AI agent skills through automated analysis and fixing iterations.
[COMMAND_EXECUTION]: Uses the Bash tool for legitimate file system operations, such as creating work directories and performing backups of target skill files before optimization. While user-provided paths are used in these commands, the context is that of a local developer tool where the user specifies the project directory to operate on.
[EXTERNAL_DOWNLOADS]: Utilizes platform-provided commands like maestro load and maestro search to retrieve coding conventions and best practice documentation. These are trusted platform-level operations intended for environment setup.
[REMOTE_CODE_EXECUTION]: Employs maestro delegate to call a trusted platform sub-agent (agy) for deep architectural and performance analysis. This is a standard pattern for multi-agent coordination within the platform.
[DATA_EXFILTRATION]: No evidence of unauthorized data transmission. The skill manages state locally in a .workflow scratchpad directory and generates human-readable reports for the user. No sensitive credentials or private files are accessed outside the target skill's scope.

skill-tuning