skills

Autonomously create a plan, consult Gemini and Codex for improvements, apply feedback, and implement. No user interaction - uses best judgment throughout.

Turn an idea into a concrete design through structured dialogue.

Gemini and Codex collaboratively brainstorm solutions, building on each other's ideas across rounds. Agent synthesizes the best ideas into a plan.

Claude brainstorms with an opponent LLM (Gemini or Codex) in alternating turns, building on each other's ideas. Synthesizes the best ideas into a plan.

Commit the staged changes. If there are no staged changes, stage all changes first.

Consult Gemini and Codex for high-level planning, synthesize into a detailed plan, implement, then get final review. No user interaction.

Consult an external LLM with the user's query.

Convert markdown to Slack-ready rich text and copy to clipboard.

Commit all pending changes as separate, logical commits using git-surgeon.

LLMs propose and critique approaches, agent moderates the debate and synthesizes the best solution, then implements.

Three-way LLM debate (Gemini, Codex, MiniMax) — each proposes, critiques the other two, then agent moderates and implements.

Orchestrates a multi-phase investigation that produces a directory of deeply-researched section documents instead of one shallow summary.

Non-interactive hunk-level git staging, unstaging, discarding, undoing, fixup, amend, squash, commit splitting, and commit reordering. Use when selectively staging, unstaging, discarding, reverting, squashing, splitting, or reordering individual diff hunks by ID instead of interactively.

Find dev servers running in tmux panes and open them in the browser. Accepts optional freetext to identify which server, or a path starting with /. Use when the user wants to open a dev server, find running servers, or open localhost.

Autonomous agent that navigates any task from start to finish. Explores the codebase, classifies complexity, adaptively consults external LLMs for planning or brainstorming, implements with incremental commits, and gets LLM review.

Stage all changes and commit directly without checking status or diff.

Rebase the current branch.

Review pending changes with external LLMs (Gemini, Codex, Claude, or all).

Walk through a multi-point topic step by step, pausing after each point to discuss with the user before continuing.

Verify review findings one by one, pausing after each to discuss with the user before continuing.

Delegate each of the proposed options to a separate AI agent in its own git worktree.

Update the current PR description based on the latest changes in context.

Manually verify review findings one by one by running code, writing adhoc scripts, and comparing actual behavior instead of implementing blindly.

Launch one or more tasks in new git worktrees using workmux.

Create an implementation plan by brainstorming with Gemini and Codex, synthesizing the best ideas, then getting their review.

Create an implementation plan for a multi-step task. Optionally review with external LLMs.

Use it when creating a bash script or editing a bash script

Use it when user asks to add images, photos, or videos to a blog post

Orchestrate multiple worktree agents -- spawn, monitor, communicate, and merge.

Claude debates an opponent LLM (Gemini or Codex) through multi-turn MCP conversation, then synthesizes the best approach and implements.

Commit, rebase, and merge the current branch.

Use it when writing or editing test files

Develop and test TUI/CLI applications by driving a tmux pane as a live test terminal. Use this skill when building interactive terminal apps (TUIs, CLIs with interactive prompts, curses/ratatui apps, REPL tools) so you can start the app, send keystrokes, capture the screen, and verify output yourself — without asking the user to test manually. Also use this when the user says things like "check it yourself", "try running it", "see if it works", or when you need to visually verify terminal output.