brainstorm
Brainstorm collaboratively with the user to develop an idea, plan, or design from scratch. Use when the user wants to explore options, ideate, or uses any 'brainstorm' trigger phrases (e.g. "let's brainstorm").
completion-check
Run a completion checklist over the current change set when work is considered done and ready to commit. Checks for correctness, simplification candidates, comment and prose audits, and repository-specific checks. Use before marking a task complete, e.g. before committing or pushing to open a PR, or when the user says "commit for me", "open a PR", "are we done".
git-worktree
Create isolated git worktrees in ~/.claude-worktrees/. Use when user says "create an isolated copy", "worktree", "create worktree", "isolated workspace", or needs to work on a branch without affecting the main working directory.
humanize-as-kevin-xiao
|
linear-fix
Autonomously completes Linear tickets via git worktrees. Use when user says "fix ENG-123", "complete LIN-456", "work on linear ticket", provides a linear.app URL, or mentions a ticket ID pattern like ABC-1234. Creates git worktree, implements fix via autonomous planning/review loop, opens GitHub PR with full reasoning.
linear-issues
Manage Linear issues - list, create, update, start/stop work, assign, comment, set priority/labels. Use when working with Linear issues, viewing tasks, creating bugs, updating issue status, adding comments, or changing assignments.
linear-search
Search Linear issues and projects by text. Use when finding issues, looking up bugs, searching the backlog, querying "what tickets mention X", or finding issues by keyword like "find ENG-" or "search for login bug".
linear-uploads
Download attachments and images from Linear issues. Use when fetching screenshots, images, videos, or file attachments from Linear comments or descriptions. Also use when user says "get the image from the ticket", "download the video", or "download the attachment".
prose-correction
Rewrite technical prose into its artifact's register and strict linear causality, then strip AI diction. Audits inline comments, docstrings, commit messages, PR titles and bodies, documentation, chat responses, or any draft text; with no target named, it audits the current change set. Use before opening a PR, during a completion check, or when the user says "prose correct this", "fix this writing", "remove the AI tells", or "make this linear".