Claude Code within tmux
Manage Claude Code instances living inside tmux sessions. Users usually create separate tmux sessions for separate projects. Use this skill when you need to read the latest Claude …
paulrahul
@paulrahul
What This Skill Does
Reads the latest Claude Code response, sends prompts, and runs /compact commands inside a named tmux session using standard tmux commands. It captures pane output, identifies the last user/assistant exchange, and sends keystrokes to the Claude pane.
Replaces manually switching to a tmux session and typing commands by letting you inspect and control Claude Code instances programmatically through tmux without helper scripts.
When to Use It
- Check the latest Claude Code response in a specific project's tmux session
- Send a new prompt to a Claude Code instance running in a tmux pane
- Trigger /compact on a Claude Code session to reduce context usage
- Retrieve the full scrollback buffer from a Claude Code pane for debugging
- Verify which tmux session contains a Claude Code pane before sending commands
Install
$ openclaw skills install @paulrahul/claude-tmuxGoal
Give Codex a repeatable checklist for interacting with Claude Code when it’s running inside tmux. Everything happens with standard tmux commands—no helper scripts. Follow these steps any time you see instructions like “check Claude in session X” or “run /compact on Claude.”
Conventions
- Session naming – We refer to tmux sessions by their tmux session name. Session names can be assigned using
tmux new-session -s <session_name>. E.g. if we had created a tmux session for project FooBar usingtmux new-session -s foobar, then we will refer to this session by the namefoobar. - Claude pane – Within a session, there should be exactly one pane whose window title or pane title is
claude. If the pane isn’t named, rename it first (Ctrl-b : select-pane -T claude). - Standard markers – Claude Code prints user prompts with
❯ …and its replies with⏺ …. We rely on that to spot the latest exchange.
Workflow
1. Locate the Claude pane
tmux list-panes -a -F '#{session_name}:#{window_index}.#{pane_index} #{pane_title}' | grep "^<session_name>" | grep -i claude
- If nothing matches, say “No pane titled ‘claude’ found inside session <name>.”
- If multiple panes match, pick the one with the lowest
window_index/pane_indexunless context says otherwise. - Record the target as
<session>:<window>.<pane>for subsequent commands.
2. View the latest exchange
tmux capture-pane -p -J -t <target> -S -200
- Scan from the bottom upward for the last
❯block (user) followed by⏺(Claude). Quote those lines back to the user. - If no
❯/⏺pair exists, say “No exchange found yet.”
3. Send a prompt
tmux send-keys -t <target> -l -- "<prompt>"
sleep 0.1
tmux send-keys -t <target> Enter
- After sending, poll using capture-pane until a new
⏺block appears (or a sensible timeout, e.g., 3 minutes). Report the reply verbatim. - If the timeout expires, say “Claude hasn’t replied yet—still waiting.”
4. Run /compact
Same as sending any prompt, but send /compact. Confirm with “Triggered /compact in session <name>.” (Claude will respond in-pane; no need to quote unless asked.)
5. Dump raw buffer (debug)
tmux capture-pane -p -J -t <target> -S -400
Use this when the user wants the entire scrollback or when parsing fails.
Tips
- Always double-check you’re addressing the right pane before sending commands—especially in shared sessions.
- If the Claude pane lives on a non-default tmux socket, prefix every tmux command with
tmux -S /path/to/socket …. - When summarizing results, mention the session/pane you used—for traceability.
- If the user wants multiple sessions handled, repeat the workflow per session.
This skill keeps things simple: pure tmux, no external code. Use it whenever you need hands-on access to Claude Code running inside tmux.
Top skills in this category
Skill Vetter
@spclaudehomeSecurity-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
Github
@steipeteInteract with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Humanizer
@biostartechnologyRemove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
Free Ride - Unlimited free AI
@shaivpidadiManages free AI models from OpenRouter for OpenClaw. Automatically ranks models by quality, configures fallbacks for rate-limit handling, and updates opencla...
Elite Longterm Memory
@nextfrontierbuildsUltimate AI agent memory system for Cursor, Claude, ChatGPT & Copilot. WAL protocol + vector search + git-notes + cloud backup. Never lose context again. Vibe-coding ready.