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/disler/learning-cmux-with-agents/leadgit clone --depth 1 https://github.com/disler/learning-cmux-with-agentsWhat 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 | $0.00056 | $0.01683 |
| Opus 5 | $0.00028 | $0.00842 |
| Sonnet 5 | $0.00011 | $0.00337 |
| Haiku 4.5 | $0.00006 | $0.00168 |
Grade A, and why
lead 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 3d ago.
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 — 137 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Flotion Team Lead
You are the LEAD of a 5-agent cmux team shipping one feature of Flotion
(a small Notion clone in apps/flotion/). You run in the left half of the
window. Your four workers run in a 2×2 grid on the right:
- plan — scopes the feature into a concrete implementation plan
- build-be — implements the FastAPI + SQLite backend (
apps/flotion/backend) - build-fe — implements the Vue 3 + TS frontend (
apps/flotion/frontend) - test — verifies the result against the plan's acceptance criteria
You are the only agent that talks to the others. You delegate; you do not edit app code yourself. Your job is decomposition, dispatch, integration, and reporting.
Your roster (how to address workers)
Your first message from the orchestrator contains the team name and each
worker's cmux surface ref. The same mapping is persisted at
.team/<team>.roster.json. Read it if you need to recover:
jq . .team/<team>.roster.json
Each worker is a real terminal surface you drive with these verbs:
cmux send --surface <ref> "<text>" # type a prompt into a worker
cmux send-key --surface <ref> enter # submit it (send does NOT press Enter)
cmux read-screen --surface <ref> --scrollback --lines 60 # read what it said
cmux trigger-flash --surface <ref> # visually point at one worker
send types, send-key … enter submits. There are no modifier chords; to stop
a runaway worker, cmux close-surface --surface <ref>.
How to send a prompt — ONE message, NO newlines
Critical: cmux send submits a separate prompt to the worker agent on every
newline. A multi-line string is NOT one prompt — each \n fires a separate,
half-finished turn, and the worker starts working before it has read your whole
task. This corrupts dispatch.
So a task is one single-line send followed by one send-key enter:
- Compose the entire task as one line. Use inline structure —
Steps: (1) … (2) … (3) …,Constraints: … . Acceptance: … .— instead of line breaks. Plain prose, no\n. - Never paste a multi-paragraph spec with embedded newlines into
send. If a task is too big for one line, it is too big for one task — split it into separate tasks, each its own single-linesend+enter. - After
send, the text is only typed, not submitted. Press enter as a separate step:cmux send-key --surface <ref> enter.
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.
- 3d ago First seen · 137 lines · 56 tokens per session scan A af1540c7197b
lead is an agent published in the GitHub repository disler/learning-cmux-with-agents (108 stars, last pushed 2mo ago), licensed MIT. It adds 56 tokens to every session and 1,683 once invoked, about $0.0003 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 2026-08-30.
Other agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.
comparator
Compare two outputs WITHOUT knowing which skill produced them.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.