Planning-first methodology using Air - a filesystem-based planning tool. Use when working with Air documents, managing project specifications, tracking implementation progress, or when the user mentions Air, airctl commands, or planning-first workflows.
Distinguished / staff+ level team role prompts for AI agents. Use when an agent needs to adopt a specific professional role — code reviewer, engineer, security analyst, UX designer, orchestrator, etc. Each role prompt sets expectations, criteria, and output format appropriate for a senior specialist.
The user's machine and shell environment on this host: OS, distro, window manager, hardware notes, default shell, and how local secret-storage tools are reached. Load before suggesting commands that depend on the shell, terminal, OS, or that need to retrieve credentials from the user's secret store.
The user's preferences for prose output style: commit messages, PR descriptions, documentation, and how the agent itself should respond. Load when writing any text intended for the user or for a repository's history.
The user's package-manager, build-tool, editor, and local-path preferences on this machine. Load before suggesting `install`, `add`, `fetch`, build commands, or invoking an editor / formatter / LSP; describes which tools are allowed, which are banned, how to invoke one-off tools, and where repos and scratch dirs live.
The user's git, review, and deployment workflow on this machine. Load before any version-control action (commit / push / merge / rebase / force-push), opening or reviewing a PR, or running a deploy.
Workarounds for using workmux with jj (Jujutsu) repositories. Use when spawning workmux agents in a jj-managed repo, whether colocated with git or native jj.
Orchestration patterns for coordinating parallel AI agents using workmux. Use when delegating tasks to multiple agents, planning parallel work, spawning workmux agents, or managing multi-agent workflows with git worktrees.