The Agent Skills Directory

[Command Execution]: The skill uses several helper scripts to manage git worktrees, tmux sessions, and external agent processes. These are essential for its function as an orchestrator and include built-in safety mechanisms such as infrastructure ledgering to ensure cleanup.
[Dynamic Context Injection]: The skill documentation describes a !command syntax used for pre-execution data injection. While this pattern allows for shell execution at load time, it is used here for legitimate development workflows (like git status) and includes specific warnings against unsafe usage like argument injection.
[Infrastructure Cleanup]: The infra_ledger.sh and lane_teardown.sh scripts implement a mandatory 'sweep' pattern. This is a security best practice that prevents resource leakage by recording every provisioned component (containers, processes, directories) and ensuring they are destroyed upon task completion.
[Human-in-the-Loop Verification]: The skill enforces a mandatory human checkpoint ('Phase 3') where no automated actions or agent launches can occur until a user provides explicit affirmative consent.

ten-lane-highway