The Agent Skills Directory

[SAFE]: The skill instructions define a legitimate process for task execution. It includes safety-oriented steps such as requiring explicit consent before working on the main branch, mandating isolation via git worktrees, and requiring a manual review of the plan for issues before execution begins.
[SAFE]: Dependencies on other 'superpowers' skills (using-git-worktrees, writing-plans, finishing-a-development-branch) appear to be part of a coordinated toolset designed for standard development workflows.
[DATA_EXPOSURE]: While the skill processes external 'plan' files (an indirect prompt injection surface), the instructions explicitly mandate a 'strict review' phase (Step 1) to identify problems or concerns in the plan before any tasks are performed, which serves as a manual mitigation strategy.

executing-plans