The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local TypeScript scripts using bun to manage the lifecycle of experiment loops. These scripts (e.g., scaffold-eval-loop.ts, manifest-sync.ts) are vendor-owned resources located within the skill's directory structure.
[COMMAND_EXECUTION]: Standard filesystem utilities such as find and sort are used to index existing artifacts in the skills-resources workspace.
[SAFE]: The skill implements a multi-agent verification chain. A Scope Guard Agent and a Critic Agent are used to validate that the loop is measurable, actionable, and respects execution boundaries, preventing the agent from performing unintended external mutations.
[SAFE]: No network operations, remote code downloads, or credential harvesting patterns were detected. All activities are confined to the local filesystem and specified project directories.

eval-loop