The Agent Skills Directory

[SAFE]: The skill utilizes the Agent tool to spawn sub-agents for plan reflection (Phase 4). These sub-agents are used strictly for quality control, reviewing BDD coverage and dependency graphs based on locally generated plan files. This is a legitimate use of the agentic framework for specialized analysis.
[SAFE]: Shell command execution is restricted to locally hosted utility scripts (docs-index.sh, seed-checklists.sh) and standard version control operations through git-agent. These actions are focused on project maintenance and documenting implementation progress.
[SAFE]: The skill ingests data from design files (bdd-specs.md, _index.md) to generate implementation plans. While this represents a surface for indirect prompt injection, the risk is managed through the skill's structured processing steps and is inherent to its primary purpose of requirements translation.

writing-plans