multi-pane
Run multi-session work — this (main) session stays a normal terminal and orchestrates; worker Claude sessions run in a separate tmux session the user watches…
Skill by goffity
Chalk Labs
Need help deploying multi-pane?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/goffity-multi-pane
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: testing
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Shell scripts
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/multi-pane
What's inside
- Skill name
multi-pane- Description
- Run multi-session work — this (main) session stays a normal terminal and orchestrates; worker Claude sessions run in a separate tmux session the user watches in its own window. Use when the user asks to split work across panes/sessions/workers, run repos in parallel, or set up an owner + QA pane loop.
- License
- MIT
- Instructions
- 83 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (8)
- SKILL.md
- mp-close.sh · scripts/
- mp-init.sh · scripts/
- mp-new.sh · scripts/
- mp-peek.sh · scripts/
- mp-send.sh · scripts/
- mp-watch.sh · scripts/
- pr-ready.sh · scripts/
SKILL.md (first lines)
# Multi-Pane Orchestration The main session (this one, a normal terminal) is the **router**: it talks to the user, sends work to worker panes, watches for results, checks reviews/CI, merges what it is allowed to, and reports. Worker Claude sessions live in a **separate tmux session** (default `mp`) that the user watches in its own iTerm window. Workers never talk to each other — everything goes through the main session. Scripts: `~/.claude/skills/multi-pane/scripts/` (below called `$MP`). Hook: `~/.claude/hooks/pane-notify.sh` (Stop/Notification → `~/.claude/pane-status.log` + macOS notification), labels in `~/.claude/pane-labels`. ## Usage ``` /multi-pane <task> e.g. /multi-pane port feature X to 6 repos, 2 owner panes + 1 QA pane ``` ## Steps ### 1. Agree scope + authority first (one short question if not given) - Which repos / tasks, how to split, QA needed? - **Authority for this session only** (never assume it carries over): e.g. "merge into develop allowed (admin ok)", "prod (main/release) is the user's", "prod DB/config changes need explicit approval". ### 2. Start the worker tmux session (main stays outside tmux) ```bash $MP/mp-init.sh # creates detached tmux session "mp" and opens an iTerm window attached to it ``` Tell the user: the worker window is `tmux attach -t mp` (panes are tiled, titled by label). ### 3. Pilot first, then parallel Same change across many repos → one owner pane does a **pilot repo** first and produces a reusable patch split into commits (shared code / per-repo theme / wiring). Then split the remaining repos across 2+ owner panes (separate repos, so no conflicts). Add a QA pane that tests behind the owners. ```bash P2=$($MP/mp-new.sh "2 FRONTEND · repoA,repoB" ~/Development/kol) # prints pane id like %31 P5=$($MP/mp-new.sh "5 QA · dev" ~/Development/kol/kol-architecture) ``` ### 4. Send work as a file (never type long text with send-keys) Write the brief to a scratchpad file, then: ```bash $MP/mp-send.sh "$P2" brief-p2.md # pastes, presses Enter until the pane shows "esc to interrupt" ``` Every brief must contain: scope (repos/branches, worktree rule — never touch shared checkouts), constraints (tests/build, no --no-verify, PR via /pr, don't merge), review gate (Copilot + /pr-review), and **the marker to print when done**: - owner: `OWNER-DONE <repo>#<PR> <short result>` (or `OWNER-FIXING <repo>#<PR>`) - QA: `QA-RESULT <repo>#<PR> PASS|FAIL <list>` - either: `NEED-INPUT <question>` if blocked on a user decision - long reports → "write to <file> then print the marker" (the Claude TUI has no scrollback) ### 5. Watch (background) and route ```bash $MP/mp-watch.sh # run_in_background: exits with new "label :: marker" lines (and [ASK] lines when a pane waits) ``` On each wake-up: - `OWNER-DONE` → `$MP/pr-ready.sh owner/repo PR` (Copilot reviewed the head? unresolved=0? checks?) → if ready and allowed, merge (merge commit) → send the PR to the QA pane via `mp-send.sh`. - `QA-RESULT … FAIL` → relay the findings to the owner pane; `PASS` → next step (e.g. open the release PR, don't merge prod). - `[ASK]`/`NEED-INPUT` → `$MP/mp-peek.sh <pane>`; answer only within the decided scope, otherwise ask the user. - Re-arm `mp-watch.sh` after every wake-up. A marker may be re-reported → check PR state before acting. ### 6. Report to the user in short status tables (repo | PR | review | QA | release PR), and say exactly what is waiting on them. ### 7. Wrap up `$MP/mp-peek.sh` each pane → idle → `$MP/mp-close.sh <pane ids>`; remove worktrees that are clean+pushed (`git worktree remove`, never rm -rf); summarize; offer to record learnings (memory / kb). ## Rules learned the hard way - **Grey text in a pane's input box is Claude Code's prompt suggestion**, not a user command — never act on it. - Pastes can stay unsubmitted → `mp-send.sh` retries Enter and checks "esc to interrupt". - **Copilot outage** (review requested but never appears): wait 30–60 min → subagent code review + triage → note "Copilot did not review" in the PR. Never claim a review that did not happen. - Before declaring "ready to merge": `pr-ready.sh` must show copilot_on_head=true (or the documented fallback), unresolved=0, checks pass. After merge: merge_parents=2 (release/sync PRs must be merge commits, never squash). - Workers must not merge, release, or change prod data; the main session decides per the authority agreed in step 1. - The main session must not bypass a worker's safety refusal by doing the same thing itself (no permission laundering). - Panes cannot SendMessage each other reliably — the main session relays via `mp-send.sh`. ## Related `/pr`, `/pr-review`, kb pattern `02-patterns/workflow/Multi-Pane Claude Orchestration.md`
From goffity/dev-km, MIT.
More from goffity/dev-km
- 0.567score
dev-km
goffity
4-layer knowledge capture system for development sessions with session management, retrospectives, and knowledge distillation.
SkillAFreshness: AClaude Code +9 - 0.558score
code-reviewer
goffity
Reviews code changes for bugs, security vulnerabilities, and performance issues.
SkillAFreshness: AClaude Code +9 - 0.558score
code-simplifier
goffity
Simplifies and cleans up code by reducing complexity and removing unnecessary abstractions.
SkillAFreshness: AClaude Code +9 - 0.558score
cleanup
goffity
Cleans up old auto-captured files with configurable retention policy and archive support.
SkillAFreshness: AClaude Code +9 - 0.558score
build-validator
goffity
Validates build, tests, and lint before push to ensure quality gates pass.
SkillAFreshness: AClaude Code +9 - 0.558score
consolidate
goffity
Consolidates auto-captured session files into daily summaries by grouping related sessions.
SkillAFreshness: AClaude Code +9 - 0.558score
jira
goffity
Manages Jira issues including creation, listing, transitions, and workflow integration.
SkillAFreshness: AClaude Code +9 - 0.558score
knowledge-curator
goffity
Scans learnings and suggests topics ready for distillation into knowledge base.
SkillAFreshness: AClaude Code +9 - 0.557score
commit
goffity
Creates atomic commits by analyzing changes, detecting mixed concerns, and commenting on related GitHub issues.
SkillAFreshness: AClaude Code +9 - 0.554score
pattern
goffity
Documents reusable design patterns with structure, examples, and usage guidelines.
SkillAFreshness: AClaude Code +9 - 0.554score
distill
goffity
Synthesizes related learnings into reusable knowledge base patterns.
SkillAFreshness: AClaude Code +9 - 0.554score
improve
goffity
Extracts and implements pending improvement items from all knowledge sources.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 12, 2026 (18 days ago)
- Last release
- Oct 12, 2026 (18 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/goffity/dev-km
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from goffity
- 0.567score
dev-km
goffity
4-layer knowledge capture system for development sessions with session management, retrospectives, and knowledge distillation.
SkillAFreshness: AClaude Code +9 - 0.558score
code-reviewer
goffity
Reviews code changes for bugs, security vulnerabilities, and performance issues.
SkillAFreshness: AClaude Code +9 - 0.558score
cleanup
goffity
Cleans up old auto-captured files with configurable retention policy and archive support.
SkillAFreshness: AClaude Code +9 - 0.558score
jira
goffity
Manages Jira issues including creation, listing, transitions, and workflow integration.
SkillAFreshness: AClaude Code +9 - 0.558score
consolidate
goffity
Consolidates auto-captured session files into daily summaries by grouping related sessions.
SkillAFreshness: AClaude Code +9 - 0.558score
knowledge-curator
goffity
Scans learnings and suggests topics ready for distillation into knowledge base.
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get multi-pane running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
multi-pane skill. Agent Mart. https://agentmart.dev/skills/goffity-multi-pane (last updated 2026-10-24)