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 skills add disler/learning-cmux-with-agents --skill cmux-architecturegit clone --depth 1 https://github.com/disler/learning-cmux-with-agentsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/disler/learning-cmux-with-agents/cmux-architecture)<a href="https://agentmods.dev/skills/disler/learning-cmux-with-agents/cmux-architecture"><img src="https://agentmods.dev/badge/skills/disler/learning-cmux-with-agents/cmux-architecture/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/disler/learning-cmux-with-agents/cmux-architecture"><img src="https://agentmods.dev/badge/skills/disler/learning-cmux-with-agents/cmux-architecture.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What 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 5.1 | $0.00063 | $0.05438 |
| Opus 5 | $0.00032 | $0.02719 |
| Sonnet 5 | $0.00013 | $0.01088 |
| Haiku 4.5 | $0.00006 | $0.00544 |
Grade A, and why
cmux-architecture 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 12d 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 — 160 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cmux Architecture
Package architecture
We are migrating cmux from a single app target into Swift Packages under Packages/. Every new package must satisfy three rules:
- Ergonomic. Public API surface matches what callers naturally want to write. Default to internal access; expose
publiconly for types and functions that downstream consumers actually use. Avoid friction such as forcing every call site through a builder or wrapper when a direct API is fine. - No dependency cycles. Packages form a strict DAG. A package may only depend on packages strictly lower in the graph. When two packages need to share a type, lift it to a common lower-level package or define a protocol seam in the consumer. Every new dependency edge requires re-checking that the graph stays acyclic.
- Clear but not overly narrow responsibilities. A package owns one full domain (e.g. settings, appearance, workspace, terminal, browser, command palette), not a slice of one. A package called "appearance math" or "workspace model" is too narrow — it forces every consumer that touches the surrounding domain to also depend on the sibling slices. Prefer a single
CmuxAppearancethat owns settings, theming, colors, glass, and snapshots together, overCmuxAppearanceMath+CmuxAppearanceTheme+CmuxAppearanceSettings. Don't fragment a domain intoCmuxFooFormatting+CmuxFooLogic+CmuxFooState— that's folder structure inside a single package, not module structure. A package boundary exists because more than one consumer needs the contents, or a build/test seam needs to exist.
When in doubt, extract leaf-first: pull out the package that has no internal dependencies. Consumers in the app target stay put and only update imports. Each leaf shrinks the app target without requiring downstream packages to exist yet.
The existing packages under Packages/ predate this policy and should not be used as design references.
Wiring a new local package into the project. cmux.xcodeproj lists package dependencies explicitly (it is not a synchronized-folder project). Adding Packages/CmuxFoo means mirroring an existing package's project.pbxproj entries — one XCLocalSwiftPackageReference (in the project's packageReferences), one XCSwiftPackageProductDependency, and a PBXBuildFile linked in the Frameworks phase of every target that imports it. The app-target packages link into both cmux and cmux-unit (so tests can import and inject them); copy a recent leaf like CmuxSocketControl for the exact shape, then run scripts/normalize-pbxproj.py and scripts/check-pbxproj.sh. A package the app builds against but cmux-unit does not link will compile the app yet fail the test target.
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 12d ago First seen · 160 lines · 63 tokens per session scan A f10498d5136e
cmux-architecture is a skill published in the GitHub repository disler/learning-cmux-with-agents (110 stars, last pushed 2mo ago), licensed MIT. It adds 63 tokens to every session and 5,438 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 skills, from other repositories
kotlin-specialist
Provides idiomatic Kotlin implementation patterns including coroutine concurrency, Flow stream handling, multiplatform architecture, Compose UI construction, Ktor server setup, and type-safe DSL design. Use when building Kotlin applications requiring coroutines, multiplatform development, or Android with Compose.…
swiftui-dev
Use this skill for SwiftUI development, architecture, structure, performance, and Apple native app profiling. It combines.
axiom-concurrency
Use when writing ANY async code, actors, threads, or seeing ANY concurrency error. Covers Swift 6 concurrency, @MainActor, Sendable, data races, async/await patterns.
wax
Swift framework guidance for Wax on-device memory/RAG. Use when writing Swift code with the public Memory facade, experimental PhotoMemory / VideoMemory, BuiltInMultimodalEmbeddings, embedding providers, retrieval modes, or hybrid search. For agent operators using the Wax MCP server tools, use the separate wax-mcp…
mobiai-kmp
Use when working on a Kotlin Multiplatform project — shared code, expect/actual declarations, platform-specific implementations, building and testing.
kotlin-concurrency-expert
Kotlin Coroutines review and remediation for Android. Use when asked to review concurrency usage, fix coroutine-related bugs, improve thread safety, or resolve lifecycle issues in Kotlin/Android code.