The Agent Skills Directory

[SAFE]: The skill implements a robust, purely local workflow for task execution and progress tracking. Analysis of the provided scripts (task-start, task-done) and the instructional markdown shows no indicators of malicious intent or security vulnerabilities. All operations are confined to the local repository and a Git-ignored workspace directory. There are no network operations, credential exposures, or attempts to escalate privileges. The use of sub-scripts from a relative path (../../subagent-driven-development/scripts) indicates integration with a larger internal framework rather than reliance on unverifiable external dependencies. The instructions emphasize safety by requiring explicit human consent for destructive operations or actions outside the worktree.

executing-plans