Menu
Verbs grouped by what you're trying to do (ask/edit/watch/run), not by capability surface
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
agent-tui usage — spawn, snapshot, wait, interact with PTY apps via @eN refs
Drive AI CLIs (opencode, pi, codex, claude-code, …) — auth, streaming, tool use, sessions
Selectors, refs, and targeted writes — the DOM-lite addressing model for nodes inside a pane
Drive common TUI apps — htop, less, lazygit, tig, fzf, nano — for engine-correctness coverage
Open files, edit, save, search in vim/nvim with adapter-aware state
Drive bash/zsh shell sessions; detect prompt vs. running command via OSC 133
| name | intent |
| description | Verbs grouped by what you're trying to do (ask/edit/watch/run), not by capability surface |
| allowed-tools | Bash(agent-tui:*) |
Pick the verb by what you're trying to accomplish. Each is sugar
over the capability primitives (spawn / wait / snapshot / …);
they exist so first contact with the tool doesn't require reading
the full surface.
This skill assumes the core verbs exist. It only documents the intent-shaped wrappers.
agent-tui ask claude "what is 40+2"
# → 42
agent-tui ask opencode "refactor this function"
agent-tui ask pi "summarize the README"
ask knows the per-CLI flags (claude → -p; opencode → run --pure --title fixed --dangerously-skip-permissions inside bash -c "cat | ..."; etc.). One verb, no incantation.
Long asks. ask defaults to a 120 s deadline (--max <ms>), and
the client read timeout is derived from --max — a long ask no longer
trips a fixed client-side cap. Raise it for very long runs:
agent-tui ask --max 300000 …. See the Timeouts note in
core for the per-verb --max defaults.
For non-AI CLIs without a recipe — or with custom flags — use
run directly:
agent-tui run --stdin "your input" -- some-cli --flags
See core for the full run surface.
agent-tui edit /work/notes.md # opens $EDITOR (default vim)
EDITOR=helix agent-tui edit /work/x.rs # override editor
agent-tui edit /work/x.rs --editor nano # explicit override
Blocks until the editor exits, then prints the file's content. Useful as a one-shot "edit and capture" without writing a wrapper.
agent-tui watch -- tail -f /var/log/syslog
agent-tui watch -- bash -c "for i in {1..10}; do echo step \$i; sleep 1; done"
Spawns the child + streams its output to stdout via tail --follow.
Exits when the child exits.
spawn + press + wait.spawn + tail --follow in one terminal, snapshot calls in
another.run --stdin directly; consider
adding a recipe upstream.