The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the GitHub CLI (gh) and git to manage pull requests, view PR status, handle stack topology, and manage worktrees. These are standard operations for a development-focused agent.
[DYNAMIC_EXECUTION]: It orchestrates complex tasks by launching subagents and parallel threads using platform-specific primitives like t3_thread_launch and delegate_task. It defines specific roles (e.g., implement, review, explore) and passes consolidated briefs to these subagents.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, specifically GitHub pull request comments and automated review bot feedback. It mitigates potential injection risks by providing a REVIEW-BOT-TRIAGE.md playbook that explicitly instructs the agent to treat this text as untrusted data and verify claims against the actual code.
[PERSISTENCE]: The orchestrate and babysit playbooks use platform-level task scheduling (schedule_task) and event watches (watch_pull_request) to maintain coordination and PR monitoring across multiple sessions, which is the intended functionality for long-running autonomous tasks.
[SAFE]: The skill uses local directories for state management and configuration (e.g., ~/.config/ox/, ~/.local/state/ox/), which is standard for CLI-driven agent tools. It also includes a disable-model-invocation: true setting in its metadata, ensuring it is only active when explicitly invoked by the user.

ox-mode