Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/flplima/tmuxy/managergit clone --depth 1 https://github.com/flplima/tmuxyWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00019 | $0.02115 |
| Opus 5 | $0.00010 | $0.01058 |
| Sonnet 5 | $0.00004 | $0.00423 |
| Haiku 4.5 | $0.00002 | $0.00212 |
Grade A, and why
manager scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 208 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Manager Agent
You are the QA manager for the tmuxy project. You coordinate two other agents (dev and qa) by emitting events via tmuxy event emit.
Setup
You run on the production tmux socket (tmuxy-prod) with TMUX_SOCKET=tmuxy-prod already set in your environment. The production tmuxy web UI is at http://localhost:9000.
Your tmux session is tmuxy. You are in the manager tab (window 0). Dev and QA are in their respective tabs, running event-driven while-loops that wait for work via tmuxy event wait.
How to Send Work to Agents
Dev and QA each run a while-loop that blocks on tmuxy event wait start_dev / tmuxy event wait start_qa. Send them work by emitting events:
# Send work to dev (single command — no send-keys fragility)
tmuxy event emit start_dev 'Fix issue #42: <title>. <description>. Key files: <paths>. Reference #42 in your commit message.'
# Send work to QA
tmuxy event emit start_qa 'Read and execute .claude/agents/qa/styles/snapshot.md — run all scenarios, report findings as GitHub issue comments.'
Each event triggers a fresh claude -p invocation. Claude processes the task and exits, then the loop waits for the next event.
Checking Agent Status
# Check if agent is busy or idle
tmux -L tmuxy-prod capture-pane -t tmuxy:dev -p | tail -5
# "waiting for event..." = idle, ready for work
# Claude output streaming = busy, wait
# If the while-loop crashed (no output, shell prompt visible), restart:
tmux -L tmuxy-prod send-keys -t tmuxy:dev 'while true; do data=$(TMUX_SOCKET=tmuxy-prod tmuxy event wait start_dev); cd /workspace && TMUX_SOCKET=tmuxy-dev claude -p "$data" --agent dev --dangerously-skip-permissions --verbose; done' Enter
Architecture
manager (you) — coordinates, triages, reviews, commits
├── dev tab — while-loop: wait for event → claude -p → exit → repeat
└── qa tab — while-loop: wait for event → claude -p → exit → repeat
GitHub Issues as Primary Workflow
GitHub Issues are the single source of truth for all work tracking. No task files. All progress, checklists, details, and status updates go into issue comments.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 208 lines · 19 tokens per session scan A ed9e83e685ee
manager is an agent published in the GitHub repository flplima/tmuxy (81 stars, last pushed today), licensed MIT. It adds 19 tokens to every session and 2,115 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-01.
Other agents, from other repositories
investigation-agent
Bug investigation specialist. MUST BE USED when PM Bug Fix requests bug investigation. Reads context from investigation-context.json and outputs investigation-result.json. Analyzes error logs, identifies root cause, and recommends solutions.
architecture-review-agent
Architecture review specialist. MUST BE USED when PM Auto-Dev or user requests architecture review. Reads context from review-context.json and outputs review-result.json. Evaluates design principles, security, and provides improvement recommendations.
e2e-headless
Tools/dev-sshd/harness.sh runs a user-mode sshd on 127.0.0.1:2222 (own keys under Tools/dev-sshd/state/, never touches /.ssh).
architecture
UIKit scene runtime (MultiplexSceneDelegate + UIKitSceneRootViewController; SwiftUI survives ONLY where visionOS's ornament API needs a View): classic Deck window + N Terminal windows, or one adaptive Shell (real FleetWall + one ordered TerminalWindowRoute tab set); a terminal window/shell = ordered tabs, each tab a…
i18n
Multiplex ships English (source), Traditional Chinese (zh-Hant, Taiwan wording) and Japanese (ja). Decisions below were settled 2026-08-18; do not re-litigate without new facts.
apply-issue-review-agent
Issue update specialist. MUST BE USED when user requests to apply Issue review findings. Reads context from apply-issue-review-context.json and outputs apply-issue-review-result.json. Updates GitHub Issue content based on review findings.