Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

cmux Architecture · manaflow-ai | DecimalAI

Codingby manaflow-ai

cmux Architecture

cmux package architecture, refactor layering, dependency inversion, file organization, DocC documentation, package design discipline, testability, and Swift 6 concurrency rules. Use before adding or meaningfully rewriting Swift files, Swift packages, coordinators, services, repositories, or public package APIs.

Never made a case worseMaintained

SkillScore

SkillScore95/100

Test results 95

Safety

3/3 safety checks passed

Code scan
Intent
Content

What it changes

45%95%of test cases passed +50 pts

Without skill10 of 22

With skill21 of 22

11 fixed10 passed either way1 still failing0 broke

Measured on gemini-3.6-flash8/4/2026v1 · latest is v2

This benchmark measured an earlier version of the skill.

A newer attempt on gemini-3.6-flash on 10/2/2026 did not produce a publishable comparison (graded differently). The snapshot above is the previously measured run.

Track record

Agent runs: 0; All time

See one task, with and without the skill

One test taskWhen adding a new macOS-only package named `MacTerminalUI`, where should its directory live physically on disk, and how should cross-group dependencies reference it?

Without the skillMissed

With the skillPassed

Same model, same task. Answers are capped and redacted.

Use this skill

Your agent runs on

Adds every file the skill has, pinned so you can update it. No account needed.

pip install "decimalai>=0.15"
decimalai skills add manaflow-ai-cmux-architecture
claude --plugin-dir "$(git rev-parse --show-toplevel 2>/dev/null || pwd)/.agents"

The same install from Python, for a notebook or a setup script.

Install

pip install "decimalai>=0.15"

Add the skill

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

print("Added. Start Claude Code with the command below.")

Start Claude Code with it

claude --plugin-dir "$(git rev-parse --show-toplevel 2>/dev/null || pwd)/.agents"

Type these in a Claude Code session. No account or API key. It writes the skill and any files it bundles, and asks first when its safety rating isn't Passed.

Add the DecimalAI plugin, once

/plugin marketplace add decimal-labs/decimalai-claude-plugins
/plugin install decimalai@decimalai

Install this skill

/decimalai:install manaflow-ai-cmux-architecture

The plugin needs python3 and curl on your machine.

No Python or account. Claude Code picks it up in its next session.

Save it into your project

curl -fsSL --create-dirs -o .claude/skills/manaflow-ai-cmux-architecture/SKILL.md https://app.decimal.ai/s/manaflow-ai-cmux-architecture/SKILL.md

Copy SKILL.md

Save it in your project, as

.claude/skills/manaflow-ai-cmux-architecture/SKILL.md

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Claude Code docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[claude-agent-sdk]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import asyncio

from claude_agent_sdk import ClaudeAgentOptions
from decimalai.claude_agent_sdk import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

async def main():
    question = "Hello!"
    async with support.run(question) as run:  # one user message, one run
        async for message in run.query(options=ClaudeAgentOptions(max_turns=6)):
            print(message)

# In a notebook or async code: await main()
asyncio.run(main())

Prefer its own loader? ClaudeAgentOptions(plugins=[skills_plugin()], skills="all") reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

The same install from Python, for a notebook or a setup script. agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[claude-agent-sdk]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import asyncio

from claude_agent_sdk import ClaudeAgentOptions
from decimalai.claude_agent_sdk import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

async def main():
    question = "Hello!"
    async with support.run(question) as run:  # one user message, one run
        async for message in run.query(options=ClaudeAgentOptions(max_turns=6)):
            print(message)

# In a notebook or async code: await main()
asyncio.run(main())

Prefer its own loader? ClaudeAgentOptions(plugins=[skills_plugin()], skills="all") reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from claude_agent_sdk import ClaudeAgentOptions, query

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
options = ClaudeAgentOptions(system_prompt=f"You are a helpful agent.\n\n{skill}", max_turns=6)

async def main():
    async for message in query(prompt="Hello!", options=options):
        print(message)

# In a notebook or async code: await main()
asyncio.run(main())

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Claude Agent SDK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[openai-agents]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from agents import Agent, Runner
from decimalai.openai_agents import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    name=support.name,
    instructions=support.instructions,  # your prompt, then the skill list
    tools=[*support.tools],             # your tools, then load_skill
)
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

Using SandboxAgent? capabilities=[Shell(), decimalai.openai_agents.sandbox_skills()] points its own loader at the same folder.

Install

pip install "decimalai[openai-agents]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from agents import Agent, Runner
from decimalai.openai_agents import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    name=support.name,
    instructions=support.instructions,  # your prompt, then the skill list
    tools=[*support.tools],             # your tools, then load_skill
)
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

Using SandboxAgent? capabilities=[Shell(), decimalai.openai_agents.sandbox_skills()] points its own loader at the same folder.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from agents import Agent, Runner

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = Agent(name="support", instructions=f"You are a helpful agent.\n\n{skill}")
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

OpenAI Agents docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[openai]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import json

from openai import OpenAI
from decimalai.openai import agent_config

client = OpenAI()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your tools, as Chat Completions takes them, and what answers each.
TOOLS = [{"type": "function", "function": {
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters": {"type": "object", "properties": {"order_id": {"type": "string"}},
                   "required": ["order_id"]},
}}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    # run.system is your prompt, then the skill list.
    messages = [{"role": "system", "content": run.system}, run.message]
    while True:
        reply = client.chat.completions.create(
            model="gpt-5.4-mini",
            messages=messages,
            tools=[*TOOLS, *run.tools],  # your tools, then load_skill
        ).choices[0].message
        messages.append(reply)
        if not reply.tool_calls:
            break
        for call in reply.tool_calls:
            # load_skill is answered by the run; your own tools by FUNCTIONS.
            function = FUNCTIONS.get(call.function.name)
            messages.append(run.tool_result(call) or {
                "role": "tool", "tool_call_id": call.id,
                "content": function(**json.loads(call.function.arguments)) if function
                else f"There is no tool named {call.function.name!r}.",
            })
print(reply.content)

# The Responses API takes run.response_tools instead of run.tools.

Install

pip install "decimalai[openai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import json

from openai import OpenAI
from decimalai.openai import agent_config

client = OpenAI()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your tools, as Chat Completions takes them, and what answers each.
TOOLS = [{"type": "function", "function": {
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters": {"type": "object", "properties": {"order_id": {"type": "string"}},
                   "required": ["order_id"]},
}}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    # run.system is your prompt, then the skill list.
    messages = [{"role": "system", "content": run.system}, run.message]
    while True:
        reply = client.chat.completions.create(
            model="gpt-5.4-mini",
            messages=messages,
            tools=[*TOOLS, *run.tools],  # your tools, then load_skill
        ).choices[0].message
        messages.append(reply)
        if not reply.tool_calls:
            break
        for call in reply.tool_calls:
            # load_skill is answered by the run; your own tools by FUNCTIONS.
            function = FUNCTIONS.get(call.function.name)
            messages.append(run.tool_result(call) or {
                "role": "tool", "tool_call_id": call.id,
                "content": function(**json.loads(call.function.arguments)) if function
                else f"There is no tool named {call.function.name!r}.",
            })
print(reply.content)

# The Responses API takes run.response_tools instead of run.tools.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from openai import OpenAI

client = OpenAI()
skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
reply = client.chat.completions.create(
    model="gpt-5.4-mini",
    messages=[
        {"role": "system", "content": f"You are a helpful agent.\n\n{skill}"},
        {"role": "user", "content": "Hello!"},
    ],
).choices[0].message

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

OpenAI SDK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[anthropic]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import anthropic
from decimalai.anthropic import agent_config

client = anthropic.Anthropic()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

question = "Hello!"
with support.run(question) as run:  # one user message, one run
    answer = client.beta.messages.tool_runner(
        model="claude-haiku-5-5",
        max_tokens=4096,
        system=run.system,     # your prompt, then the skill list
        tools=[*run.tools],    # your tools, then load_skill
        messages=[{"role": "user", "content": question}],
    ).until_done()

Install

pip install "decimalai[anthropic]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import anthropic
from decimalai.anthropic import agent_config

client = anthropic.Anthropic()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

question = "Hello!"
with support.run(question) as run:  # one user message, one run
    answer = client.beta.messages.tool_runner(
        model="claude-haiku-5-5",
        max_tokens=4096,
        system=run.system,     # your prompt, then the skill list
        tools=[*run.tools],    # your tools, then load_skill
        messages=[{"role": "user", "content": question}],
    ).until_done()

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
import anthropic

client = anthropic.Anthropic()
skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
reply = client.messages.create(
    model="claude-haiku-5-5",
    max_tokens=4096,
    system=f"You are a helpful agent.\n\n{skill}",
    messages=[{"role": "user", "content": "Hello!"}],
)

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Anthropic docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[google-genai]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from google import genai
from google.genai import types
from decimalai.google_genai import agent_config

client = genai.Client()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your functions, as declarations for the model, and what answers each.
TOOLS = [{"function_declarations": [{
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters_json_schema": {"type": "object", "properties": {"order_id": {"type": "string"}},
                               "required": ["order_id"]},
}]}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    config = types.GenerateContentConfig(
        system_instruction=run.system,     # your prompt, then the skill list
        tools=[*TOOLS, *run.tool_params],  # your functions, then load_skill
    )
    contents = [{"role": "user", "parts": [{"text": run.message}]}]
    while True:
        response = client.models.generate_content(
            model="gemini-3.5-flash", contents=contents, config=config,
        )
        if not response.function_calls:
            break
        contents.append(response.candidates[0].content)
        parts = []
        for call in response.function_calls:
            # load_skill is answered by the run; your own functions by FUNCTIONS.
            # A loop of your own rather than automatic function calling, which
            # raises KeyError when the model asks for a function you did not pass.
            function = FUNCTIONS.get(call.name)
            parts.append(run.tool_result(call) or {"function_response": {
                "name": call.name, "id": call.id,
                "response": {"result": function(**(call.args or {})) if function
                             else f"There is no function named {call.name!r}."},
            }})
        contents.append({"role": "user", "parts": parts})
print(response.text)

Install

pip install "decimalai[google-genai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from google import genai
from google.genai import types
from decimalai.google_genai import agent_config

client = genai.Client()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your functions, as declarations for the model, and what answers each.
TOOLS = [{"function_declarations": [{
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters_json_schema": {"type": "object", "properties": {"order_id": {"type": "string"}},
                               "required": ["order_id"]},
}]}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    config = types.GenerateContentConfig(
        system_instruction=run.system,     # your prompt, then the skill list
        tools=[*TOOLS, *run.tool_params],  # your functions, then load_skill
    )
    contents = [{"role": "user", "parts": [{"text": run.message}]}]
    while True:
        response = client.models.generate_content(
            model="gemini-3.5-flash", contents=contents, config=config,
        )
        if not response.function_calls:
            break
        contents.append(response.candidates[0].content)
        parts = []
        for call in response.function_calls:
            # load_skill is answered by the run; your own functions by FUNCTIONS.
            # A loop of your own rather than automatic function calling, which
            # raises KeyError when the model asks for a function you did not pass.
            function = FUNCTIONS.get(call.name)
            parts.append(run.tool_result(call) or {"function_response": {
                "name": call.name, "id": call.id,
                "response": {"result": function(**(call.args or {})) if function
                             else f"There is no function named {call.name!r}."},
            }})
        contents.append({"role": "user", "parts": parts})
print(response.text)

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from google import genai
from google.genai import types

client = genai.Client()
skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
response = client.models.generate_content(
    model="gemini-3.5-flash",
    contents="Hello!",
    config=types.GenerateContentConfig(system_instruction=f"You are a helpful agent.\n\n{skill}"),
)

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Gemini API docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[langchain]" langchain-openai
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from langchain.agents import create_agent
from decimalai.langchain import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = create_agent(
    "openai:gpt-5.4-mini",
    tools=[],                         # your tools; the middleware adds load_skill
    middleware=[support.middleware],  # your prompt, the skill list and load_skill
)
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

Install

pip install "decimalai[langchain]" langchain-openai

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from langchain.agents import create_agent
from decimalai.langchain import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = create_agent(
    "openai:gpt-5.4-mini",
    tools=[],                         # your tools; the middleware adds load_skill
    middleware=[support.middleware],  # your prompt, the skill list and load_skill
)
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from langchain.agents import create_agent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = create_agent("openai:gpt-5.4-mini", tools=[], system_prompt=f"You are a helpful agent.\n\n{skill}")
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

LangChain docs

CrewAI's own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" "crewai>=1.15.27"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import decimalai.crewai
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                               # your own CrewAI LLM
    skills=decimalai.crewai.skills(),      # the installed folder
    tools=decimalai.crewai.skill_tools(),  # reads a skill's files, runs its scripts
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

The same install from Python, for a notebook or a setup script. CrewAI's own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" "crewai>=1.15.27"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import decimalai.crewai
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                               # your own CrewAI LLM
    skills=decimalai.crewai.skills(),      # the installed folder
    tools=decimalai.crewai.skill_tools(),  # reads a skill's files, runs its scripts
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

No DecimalAI package. CrewAI's own loader reads the folder, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                         # your own CrewAI LLM
    skills=[Path(".agents/skills")],
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

CrewAI docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[adk]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from decimalai.adk import DecimalAIPlugin, agent_config
from google.adk.agents import LlmAgent
from google.adk.apps import App

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = LlmAgent(
    name="support",          # ADK needs a Python identifier
    model="gemini-3.5-flash",
    instruction=support.instruction,  # your prompt, then the skill list
    tools=[*support.tools],           # your tools, then load_skill
)
# DecimalAIPlugin makes each invocation of runners built from this App one trace.
app = App(name="support", root_agent=agent, plugins=[DecimalAIPlugin()])

Prefer ADK's own loader? SkillToolset(skills=decimalai.adk.skills(), ...) reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

Install

pip install "decimalai[adk]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from decimalai.adk import DecimalAIPlugin, agent_config
from google.adk.agents import LlmAgent
from google.adk.apps import App

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = LlmAgent(
    name="support",          # ADK needs a Python identifier
    model="gemini-3.5-flash",
    instruction=support.instruction,  # your prompt, then the skill list
    tools=[*support.tools],           # your tools, then load_skill
)
# DecimalAIPlugin makes each invocation of runners built from this App one trace.
app = App(name="support", root_agent=agent, plugins=[DecimalAIPlugin()])

Prefer ADK's own loader? SkillToolset(skills=decimalai.adk.skills(), ...) reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from google.adk.agents import LlmAgent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = LlmAgent(
    name="support",
    model="gemini-3.5-flash",
    instruction=lambda context: f"You are a helpful agent.\n\n{skill}",
)

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Google ADK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[pydantic-ai]"
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from pydantic_ai import Agent
from decimalai.pydantic_ai import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    "openai:gpt-4o-mini",
    name=support.name,
    tools=[],                 # your tools; the capability adds load_skill
    capabilities=[support],   # your prompt, the skill list and load_skill
)
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

On the Pydantic AI harness, decimalai.pydantic_ai_harness.skills(project) uses its own loader instead.

Install

pip install "decimalai[pydantic-ai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from pydantic_ai import Agent
from decimalai.pydantic_ai import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    "openai:gpt-4o-mini",
    name=support.name,
    tools=[],                 # your tools; the capability adds load_skill
    capabilities=[support],   # your prompt, the skill list and load_skill
)
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

On the Pydantic AI harness, decimalai.pydantic_ai_harness.skills(project) uses its own loader instead.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from pydantic_ai import Agent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = Agent("openai:gpt-4o-mini", instructions=f"You are a helpful agent.\n\n{skill}")
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Pydantic AI docs

Deep Agents' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" deepagents
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import decimalai.deepagents
from deepagents import create_deep_agent

agent = create_deep_agent(model=model, **decimalai.deepagents.skills())
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

The default backend reads files but has no shell. To run a skill's scripts, pass skills(backend=LocalShellBackend(...)).

The same install from Python, for a notebook or a setup script. Deep Agents' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" deepagents

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import decimalai.deepagents
from deepagents import create_deep_agent

agent = create_deep_agent(model=model, **decimalai.deepagents.skills())
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

The default backend reads files but has no shell. To run a skill's scripts, pass skills(backend=LocalShellBackend(...)).

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from deepagents import create_deep_agent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = create_deep_agent(model=model, system_prompt=f"You are a helpful agent.\n\n{skill}")
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Deep Agents docs

Its own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" agent-framework
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

import asyncio

from agent_framework import Agent
from decimalai.agent_framework import skills

agent = Agent(client=client, instructions="You are a helpful agent.", context_providers=[skills()])
asyncio.run(agent.run("Hello!"))

It caches skills by default, so restart the process after decimalai skills update.

The same install from Python, for a notebook or a setup script. Its own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" agent-framework

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

import asyncio

from agent_framework import Agent
from decimalai.agent_framework import skills

agent = Agent(client=client, instructions="You are a helpful agent.", context_providers=[skills()])
asyncio.run(agent.run("Hello!"))

It caches skills by default, so restart the process after decimalai skills update.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from agent_framework import Agent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = Agent(client=client, instructions=f"You are a helpful agent.\n\n{skill}")
asyncio.run(agent.run("Hello!"))

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Microsoft Agent Framework docs

Strands' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" strands-agents
decimalai skills add manaflow-ai-cmux-architecture

Add it to your agent

from decimalai.strands import skills
from strands import Agent
from strands.vended_tools import file_editor, shell

agent = Agent(
    model=model,                # always pass one: the default is Amazon Bedrock
    system_prompt="You are a helpful agent.",
    tools=[file_editor, shell],
    plugins=[skills()],
)
agent("Hello!")

Strands loads an agent's skills once, so build the agent again after decimalai skills update.

The same install from Python, for a notebook or a setup script. Strands' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" strands-agents

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("manaflow-ai-cmux-architecture")
if result.exit_code:
    raise SystemExit(result.problems)

from decimalai.strands import skills
from strands import Agent
from strands.vended_tools import file_editor, shell

agent = Agent(
    model=model,                # always pass one: the default is Amazon Bedrock
    system_prompt="You are a helpful agent.",
    tools=[file_editor, shell],
    plugins=[skills()],
)
agent("Hello!")

Strands loads an agent's skills once, so build the agent again after decimalai skills update.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

Give it to your agent

from pathlib import Path
from strands import Agent

skill = Path(".agents/skills/manaflow-ai-cmux-architecture/SKILL.md").read_text()
agent = Agent(model=model, system_prompt=f"You are a helpful agent.\n\n{skill}")
agent("Hello!")

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Strands docs

Adds every file the skill has, pinned so you can update it. Cursor, Copilot, Cline and Warp read this folder by themselves.

pip install "decimalai>=0.15"
decimalai skills add manaflow-ai-cmux-architecture

No Python or account. Cursor, Copilot, Cline and Warp read this folder by themselves.

Save it into your project

curl -fsSL --create-dirs -o .agents/skills/manaflow-ai-cmux-architecture/SKILL.md https://app.decimal.ai/s/manaflow-ai-cmux-architecture/SKILL.md

Copy SKILL.md

Save it in your project, as

.agents/skills/manaflow-ai-cmux-architecture/SKILL.md

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Cursor and other editors docs

Fetch the file and give its text to your model as instructions. No account or package needed.

Get the file

curl -fsSL https://app.decimal.ai/s/manaflow-ai-cmux-architecture/SKILL.md

Read the summary as JSON

curl -fsSL https://app.decimal.ai/s/manaflow-ai-cmux-architecture.json

This gets SKILL.md only. The skill's 4 other files are not included.

Other ways to add it

Other setups docs

You don't need a DecimalAI account.

Version: v2 (v1 tested 8/4/2026)
Source: manaflow-ai/cmux ↗
License: NOASSERTION

Manage it in DecimalAI

Saved prompts and skills, pinned versions, and traces of this skill in your runs. DecimalAI does not run your agent.

Claude Code reads files from your project, so DecimalAI config does not apply there.

On Claude Agent SDK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On OpenAI Agents: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On OpenAI SDK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Anthropic: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Gemini API: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On LangChain: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On CrewAI: set DECIMAL_API_KEY and add decimalai.init(crewai=True) to record runs and skill use. The skill version stays in your repo.

On Google ADK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Pydantic AI: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Deep Agents: set DECIMAL_API_KEY and add decimalai.init(langchain=True) to record runs and skill use. The skill version stays in your repo.

On Microsoft Agent Framework: set DECIMAL_API_KEY and add decimalai.init(otel=True, agent_name="support") to record runs and skill use. The skill version stays in your repo.

On Strands: set DECIMAL_API_KEY and add decimalai.init(otel=True, agent_name="support") to record runs and skill use. The skill version stays in your repo.

Editors read files from your project, so DecimalAI config does not apply there.

Anywhere else: record a use with decimalai.log_skill_activation() inside a trace.

Use this skill

SKILL.md

cmux Architecture

Package architecture

cmux is migrating from a single app target into Swift Packages under Packages/. Every new package must be:

Ergonomic. Default to internal access; public only what downstream consumers actually use.
Acyclic. Packages form a strict DAG. Share a type by lifting it to a lower package or defining a protocol seam in the consumer. Every new dependency edge requires re-checking that the graph stays acyclic.
Whole-domain. One package owns a full domain (settings, appearance, workspace, terminal, browser, command palette). CmuxAppearanceMath + CmuxAppearanceTheme + CmuxAppearanceSettings is folder structure inside CmuxAppearance, not module structure. A boundary exists because more than one consumer needs the contents, or a build/test seam must exist.

When in doubt, extract leaf-first: the package with no internal dependencies. Existing packages under Packages/ predate this policy; do not use them as design references.

Wiring a new package into cmux.xcodeproj needs explicit pbxproj entries in both the cmux and cmuxTests targets (cmuxTests is what the cmux-unit scheme runs). See references/package-boundaries.md.

Group folders. Every package lives physically under exactly one group directory: Packages/Shared/<pkg> (both apps), Packages/iOS/<pkg> (iOS only), or Packages/macOS/<pkg> (macOS only). cmux.xcworkspace/contents.xcworkspacedata mirrors that folder shape, with three groups whose container locations are those folders and every package directory as a FileRef under its folder's group. The folder is the source of truth: to move a package, git mv the directory then run python3 scripts/check-workspace-package-groups.py --write. Cross-group .package(path:) deps use ../../<Group>/<Name>. Never hand-edit workspace group membership. CI runs python3 scripts/check-workspace-package-groups.py --check and fails on drift.

Lockfiles. Do not gitignore cmux-owned Package.resolved files; SwiftPM resolution changes must be visible in PR diffs. Track the root Xcode lockfile and every cmux-owned package-local Package.resolved produced by standalone swift package resolve / swift build / swift test. A package-local lockfile is the source of truth for that package's standalone resolution and is not replaced by cmux.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved. Vendored third-party directories may keep their upstream ignore policy. CI runs python3 scripts/check-package-resolved-policy.py.

Feature flags mean remote PostHog runtime flags. Unless the user explicitly asks for a compile-time flag, local setting, or environment variable, implement a feature flag through CmuxFeatureFlags with a PostHog key, an explicit unavailable fallback, registry metadata, live update behavior, and focused tests. A local override may support dogfood but must not be the production control plane.

Layers

Five layers, dependencies point only downward:

Core (CmuxCore): pure Sendable values, IDs, DTOs, errors, shared protocol seams. No AppKit/SwiftUI/I/O. The lift target when two domains need the same type.
Services / infrastructure: actors implementing core protocols against the outside world (process/PTY, filesystem, sockets, web API, notifications, auth). One package per cohesive capability.
Domain / state: @MainActor @Observable models plus Coordinators, one package per feature domain, owning that domain's mutable state. Exemplar CmuxSettings.
UI: SwiftUI/AppKit views, one UI package per domain package, depending only on its domain package plus Core, never a Service directly. Exemplar CmuxSettingsUI.
Executable (cmuxApp / AppDelegate): thin composition shim, no business logic.

Classify every extracted entity by intent:

Coordinator: @MainActor @Observable orchestrator that sequences a user flow and owns navigation/selection/lifecycle state, calling Services and child models. Does no I/O itself.
Service: actor (or @MainActor only when an AppKit main-thread API forces it) performing one outside-world capability; exposes async/await plus AsyncStream; holds only its own resource handles and no UI state.
Repository: actor mediating one persistence source of truth (file, defaults, web API) behind CRUD-shaped async methods returning value types. Precedents: JSONConfigStore, UserDefaultsSettingsStore.

Dependency inversion. Lower packages publish protocols; concrete Services/Repositories conform; higher layers depend on any Protocol, never the concrete type, and never a stored property reaching across modules. Constructor (init) injection only: no global container, no singleton, no static let shared. The executable app target is the single composition root, the one place concretes are named and the object graph is assembled. SwiftUI Environment may carry already-constructed @Observable models down a view tree (as SettingsRuntime does), never service wiring.

State and SwiftUI. Domain state lives in @MainActor @Observable models, never ObservableObject/@Published. A god model decomposes into cohesive child @Observable sub-models owned by their domain packages and composed by held reference; cross-domain reads go behind read-only protocols. In views use @State (owned), @Bindable or plain let (passed in), or @Environment(M.self) plus .environment(...) (injected). Never @StateObject / @ObservedObject / @EnvironmentObject / .environmentObject(_:).

Executable-target boundary (invert, never work around):

@main cmuxApp and AppDelegate stay in the executable target as the thin composition shim. That residual is the intended end state, not debt.
A type is declared in exactly one module and a lower package cannot extend a higher-owned type, so AppDelegate+* / cmuxApp+* / Workspace+* extensions do not move down. Extract the behavior into a Coordinator/Service/Repository, have the god object own an instance, and reduce the extension to a one-line forward.
Stored properties cannot cross module boundaries. Decompose god-model state into child @Observable sub-models owned by domain packages, composed by held reference, with cross-cutting reads behind read-only protocols.

File organization

One major type per file, named after the type (Control.swift, LabeledChoice.swift, ListControl.swift, not one shared SettingControl.swift). Applies to all new code in Packages/ and all new app-target files.

Trivial private helpers, nested types, and single-line extensions used only inside the file may stay with the parent type. Anything with a meaningful body gets its own file, including a private final class nested in another file's type.
Conformance-adding extensions for a type defined elsewhere go in TypeName+Conformance.swift or TypeName+Feature.swift, not bundled into the consuming feature file.
Type-erased wrappers live next to what they erase: Foo.swift and AnyFoo.swift.
The god files (ContentView.swift, Workspace.swift, TabManager.swift, cmuxApp.swift) are what this rule exists to stop. Splitting one file per type is correct even if it triples the file count. File count is cheap; "find this type" being unanswerable is expensive.

Documentation

Every public symbol in a new package under Packages/ gets a Swift-DocC /// comment at the time it is written. Docs are part of the API surface, not follow-up work.

First line is a one-sentence summary that fits on one line and ends with a period. Blank /// line before any discussion paragraph. Use - Parameter name: / - Returns: / - Throws: on init and func symbols that take parameters or throw. Markdown is fine.
Reference symbols with double backticks ( CmuxSetting ); plain backticks for non-symbol code (UserDefaults.standard).
Document what a type represents and when to use it, the meaning of each enum case, init parameter defaults and their reason, property invariants, method behavior, and which generic Value/Element shapes are accepted and why.
Non-trivial APIs get at least one short example in a fenced swift block, ideally a real declaration from this codebase.
internal and private symbols get a one-line /// when the intent is non-obvious. The public boundary is the one that needs full coverage.
Update the doc comment in the same edit that changes behavior or signature. Doc comments describe the contract from outside; inline // is reserved for non-obvious why.

Main app target code is not retroactively required to be documented.

Package design discipline

No shared-singleton accessors. static let standard / shared / default on a package type holding runtime state is a singleton by another name. Construct at the app startup site and inject. static let is fine for declarations (identifiers, schema entries, enum cases), not for behavior.
No namespace-enums. enum Foo { static func bar() } is a fake namespace with no instances, no DI, and no test seam. Prefer a value-typed struct passed via constructor when the helper might gain configuration.
No parallel hand-maintained registries. When a list mirrors declared items (catalog.all mirroring stored properties), derive it via Mirror reflection or a macro. Two sources of truth drift silently.
Prefer compile-time invariants to runtime traps. guard ... else { assertionFailure(...); return default } for a "programmer error" case should be encoded in the type system (phantom types, separate concrete flavors). Runtime traps become silent fallbacks in release builds.
No free functions. Top-level func declarations, any visibility including file-scope private func, are banned; scope functionality to the entity that owns the responsibility. The only sanctioned exception is a @convention(c) trampoline a C API forces, with a one-line justification.

Testability

Every public type added to Packages/ must be testable from a test target without launching the app target, booting AppKit, or depending on the user's filesystem or UserDefaults.standard.

UserDefaults, FileManager, on-disk paths, environment variables, and clocks arrive through init parameters. Tests pass a UserDefaults(suiteName:) scoped to the test, a temp directory URL, a fixed Date.
No implementation hardcodes .shared / .standard.
No static test hooks. A nonisolated(unsafe) static var fooForTesting (or any global mutable override a test swaps in) leaks across tests and usually needs a lock. Replace it with a protocol seam through init, e.g. init(commandRunner: any CommandRunning = CommandRunner()). Deleting the static hook and its lock is part of the extraction, not a follow-up.
Prefer returning the changed value and letting the caller persist over mutating global state and returning Void.
Surface observation as AsyncStream so tests can assert the yielded sequence, rather than NotificationCenter-only patterns that need a runloop spin.
Show the test-instantiation pattern in the package README.md or DocC catalog.

If a design is hard to test, it is wrong. Reach for the constructor parameter list, not the test bench.

Swift 6 concurrency

New code in Packages/, new app-target files, and meaningful rewrites use actor, async/await, AsyncStream/AsyncSequence, @Observable, and @MainActor.

Forbidden without a written justification in the PR description:

Locks: NSLock, NSRecursiveLock, os_unfair_lock, OSAllocatedUnfairLock, pthread_mutex_t, Synchronization.Mutex, DispatchSemaphore used as a lock. Ongoing mutable shared state belongs in an actor with async reads and writes.
KVO via NSObject subclassing to override observeValue(forKeyPath:...) or call addObserver(_:forKeyPath:...). Use NotificationCenter.default.notifications(named:) or the NSKeyValueObservation token API at the seam only.
DispatchQueue as a synchronization primitive (queue.sync { ... } serializing mutable state). Queues are fine for event delivery, not for protecting state.
Combine for change propagation: @Published, ObservableObject, PassthroughSubject/CurrentValueSubject, AnyCancellable.
Completion-handler public APIs ((Result<T, Error>) -> Void, (T?, Error?) -> Void). Use async throws -> T; wrap a legacy callback with withCheckedContinuation/withCheckedThrowingContinuation confined to that one seam.
DispatchQueue.main.async. Annotate the destination @MainActor and await it.
Sleeping as a synchronization substitute: any sleep used to poll for a condition, let state settle before reading, or race a callback/animation. DispatchQueue.asyncAfter is banned outright (not cancellable by structure, not testable).
A single-method actor used as a mutex. actor Guard { func claim() -> Bool } forces synchronous callers (a Process termination handler, a DispatchSource event handler, a withCheckedContinuation resume race) through Task { await guard.claim() }, adding suspension points, ordering hops, and reentrancy surface to a fundamentally synchronous compare-and-set. Use the lock carve-out instead.

Required shape: mutable shared state to an actor with async accessors and an AsyncStream for observers; SwiftUI-render-friendly state to an @Observable @MainActor view model subscribing to that stream and projecting snapshots (never read actor state synchronously from view code); cross-process and cross-thread invariants expressed through actor isolation; new public observable surfaces as AsyncStream/AsyncSequence.

Carve-outs, each with a one-line justification comment on the declaration and hidden behind an AsyncStream or actor surface so callers never see them:

DispatchSource.makeFileSystemObjectSource for file watching, makeReadSource/makeWriteSource for low-level socket I/O (no async-native replacement).
A bounded, cancellable Clock.sleep (preferred) or Task.sleep for a genuine delay or deadline that is itself the intended behavior (minimum display duration, auto-dismiss, check timeout). Drive it from an injected Clock so tests advance virtual time, store the Task, and cancel it on the relevant lifecycle transition. Never to poll, settle, or race.
DispatchSource.makeTimerSource (one-shot) only when a genuine deadline must fire outside any async context, in a non-async type with no Task to host the sleep. Prefer Clock.sleep whenever the code is already async or actor-isolated; a raw timer is not cancellation-integrated or testable and has suspend/resume/cancel footguns.
A lock for a short, non-blocking synchronous compare-and-set called from non-async callbacks. Canonical case: several synchronous Process/DispatchSource callbacks race to resume one withCheckedContinuation exactly once, guarded by OSAllocatedUnfairLock(initialState:) over a Bool. Not for guarding ongoing domain state.
NSKeyValueObservation token when wrapping a Foundation/AppKit type that exposes change only via KVO.

When extracting existing code that uses a forbidden primitive, reshape it at the seam instead of copying it; usually it wants an actor. Drain Process pipes concurrently on detached tasks keyed by the raw fd (Int32 is Sendable, FileHandle is not).

@unchecked Sendable and nonisolated(unsafe) require a comment explaining the safety argument, or the diff is rejected. @unchecked Sendable on an entire actor or struct is almost always wrong; prefer nonisolated(unsafe) let on the single non-Sendable property.

Existing app-target code may keep the old primitives until rewritten. Do not retrofit blindly.

Code linked into the macOS app also builds with Xcode 16.2 on Intel, so it stays within Swift 6.0 syntax; see Swift 6.0 compatibility.

Detailed references

references/package-boundaries.md: extraction order, dependency graph, composition root, pbxproj wiring.
references/concurrency-carveouts.md: carve-out examples and the reviewer reject-list.
references/file-api-discipline.md: one-type-per-file, DocC, design smells.
references/swift-6-0-compatibility.md: Swift 6.0 syntax limits for app-linked code (Intel/Xcode 16.2 pathway).

About this skill

Automatic

Coding pack

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v5 — 10/1/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

manaflow-ai / cmux

skills/cmux-architecture/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 10/1/2026Version v2Bundled 4 files

Illustrative Want to see what manaflow-ai/cmux-architecture would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Related skills

Skills that do similar work, with their headline benchmark lift.

From the same source and nearby in the catalogue.

Ready to supercharge your AI agents?

SkillScore

Safety

What it changes

Track record

Use this skill

cmux Architecture

Package architecture

Layers

File organization

Documentation

Package design discipline

Testability

Swift 6 concurrency

Detailed references

SkillSafety

Source

Related skills

More in the registry