### Install Skill from Local Directory

Source: https://cosine.sh/docs/customizing/skills

Installs a skill from a local filesystem path.

```bash
# Install from a local directory
cos skills add ./path/to/skills-repo --skill my-skill
```

--------------------------------

### Deploy Static Site Folder

Source: https://cosine.sh/docs/customizing/instant-sites

Uploads a static site folder and creates a preview deployment. If the project is not configured, it can guide through setup.

```bash
cos deploy <folder>
```

--------------------------------

### Install a Plugin using Shorthand Notations

Source: https://cosine.sh/docs/customizing/plugins

Shorthand formats for GitHub repositories or domain-prefixed URLs can also be used for plugin installation.

```bash
# GitHub shorthand
cos plugin install user/my-plugin
```

```bash
# Domain-prefixed
cos plugin install github.com/user/my-plugin
```

--------------------------------

### Install Skill using GitHub Shorthand

Source: https://cosine.sh/docs/customizing/skills

Installs a skill using a concise GitHub owner/repo format.

```bash
# GitHub shorthand also works
cos skills add anthropics/skills --skill skill-creator
```

--------------------------------

### Show All Install Script Options

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Display all available options and flags for the Cosine CLI one-line installation script using the --help flag.

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --help
```

--------------------------------

### Install a Plugin from a Git Repository

Source: https://cosine.sh/docs/customizing/plugins

Use this command to install a plugin by providing its full git repository URL.

```bash
cos plugin install https://github.com/user/my-plugin.git
```

--------------------------------

### List Installed Plugins

Source: https://cosine.sh/docs/customizing/plugins

View a list of all installed plugins, their versions, descriptions, and installation paths.

```bash
cos plugin list
```

--------------------------------

### Install Multiple Skills from a Repository

Source: https://cosine.sh/docs/customizing/skills

Installs selected skills from a larger repository by repeating the --skill flag.

```bash
cos skills add anthropics/skills --skill skill-creator --skill pdf
```

--------------------------------

### Install Skill from GitHub Repository

Source: https://cosine.sh/docs/customizing/skills

Installs a specific skill from a GitHub repository using its full URL.

```bash
# Install a specific skill from a GitHub repo
cos skills add https://github.com/anthropics/skills --skill skill-creator
```

--------------------------------

### Good Context-First Prompt Example

Source: https://cosine.sh/docs/learn/plan-it-first

This example demonstrates effective context-first prompting by providing all necessary details in a single message, leading to better agent understanding and execution.

```text
Add a "Generate Report" button to the analytics dashboard at
/src/pages/dashboard.tsx.


Follow the pattern used in /src/pages/settings.tsx for the
"Update Profile" button — same styling, same modal behavior.


The modal should contain a simple form with:
- Date range picker (use the existing DatePicker component)
- Report type dropdown (options: summary, detailed, export)
- Generate button that POSTs to /api/reports


Styling should match our design system — see /src/styles/components.md
for color tokens and spacing rules.


The feature is for marketing managers who need to pull weekly
performance data without engineering help.

```

--------------------------------

### Install CLI with Winget (Windows)

Source: https://cosine.sh/docs/cli/quickstart

Use this command to install the Cosine CLI on Windows using Winget. Verify the installation by checking the version.

```bash
winget install Cosine.CLI
cos --version
```

--------------------------------

### Install CLI with curl (Linux Alternative)

Source: https://cosine.sh/docs/cli/quickstart

Use this command to install the Cosine CLI on Linux by downloading and executing the installation script. Verify the installation by checking the version.

```bash
curl -fsSL https://cosine.sh/install | bash
cos --version
```

--------------------------------

### Start Agent with Prompt from File

Source: https://cosine.sh/docs/cli/changelog

Use the `--prompt-file` flag to specify a prompt from a file when starting the agent. This is useful for longer or templated prompts.

```bash
cos2 start --prompt-file ./task.md

```

--------------------------------

### Project-Scoped Skill Installation

Source: https://cosine.sh/docs/customizing/skills

Installs a skill to the project's managed directory, best for project-specific skills.

```bash
cos skills add anthropics/skills --skill skill-creator --project
```

--------------------------------

### File System Operation Example

Source: https://cosine.sh/docs/mcp/server-management

Example interaction where the user requests to read a file, and Cosine CLI uses the `mcp_filesystem` tool to perform the operation.

```text
You: Read the README.md file in my documents folder
Cosine: I'll read that file for you. [Uses mcp_filesystem tool]
```

--------------------------------

### Python MCP Server Example

Source: https://cosine.sh/docs/mcp/server-management

A basic example of an MCP server written in Python. It defines a single tool 'search_docs' and runs the server.

```python
from mcp.server import Server
from mcp.types import TextContent


app = Server("my-server")


@app.tool()
def search_docs(query: str) -> str:
    """Search documentation for a query."""
    # Your implementation here
    return f"Results for: {query}"


if __name__ == "__main__":
    app.run()

```

--------------------------------

### Global Skill Installation

Source: https://cosine.sh/docs/customizing/skills

Installs a skill to the global managed directory, suitable for personal skills used across sessions.

```bash
cos skills add anthropics/skills --skill skill-creator --global
```

--------------------------------

### Poor Context-Dribbling Prompt Example

Source: https://cosine.sh/docs/learn/plan-it-first

This example shows a common mistake of providing instructions incrementally, which leads to less effective agent performance.

```text
> Add a button to the dashboard
> Actually make it blue
> And it should open a modal
> The modal should have a form
> Use the same pattern as the user settings page

```

--------------------------------

### Start with a Specific Profile

Source: https://cosine.sh/docs/cli/config

Use the `--profile` flag with the `cos start` command to activate a named configuration profile. This allows for distinct settings for different environments like work or personal use.

```bash
# Use work profile
cos --profile work start


# Use personal profile
cos --profile personal start
```

--------------------------------

### Install Cosine CLI via Direct Download (Linux)

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Install the Cosine CLI on Linux by downloading the binary directly. Extract the archive and move the binary to a directory in your PATH.

```bash
unzip cos-linux-amd64.zip
sudo mv cos /usr/local/bin/
cos --version
```

--------------------------------

### Basic Skill Installation Command

Source: https://cosine.sh/docs/customizing/skills

The fundamental command to add a skill from a specified source.

```bash
cos skills add <source>
```

--------------------------------

### Start Chrome with Remote Debugging

Source: https://cosine.sh/docs/cli/config/flags

Instructions for starting Chrome with remote debugging enabled, necessary for using the `--cdp-url` flag with local or remote Chrome instances.

```bash
# macOS
open -n -a "Google Chrome" --args --remote-debugging-port=9222 --user-data-dir=/tmp/cosine-chrome

```

```bash
# Linux
google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/cosine-chrome

```

--------------------------------

### TypeScript MCP Server Example

Source: https://cosine.sh/docs/mcp/server-management

A basic example of an MCP server written in TypeScript. It sets up a server, defines a request handler for a 'search_docs' tool, and connects using stdio transport.

```typescript
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";


const server = new Server(
  {
    name: "my-server",
    version: "1.0.0",
  },
  {
    capabilities: {
      tools: {},
    },
  },
);


server.setRequestHandler(CallToolRequestSchema, async (request) => {
  if (request.params.name === "search_docs") {
    const { query } = request.params.arguments;
    return {
      content: [{ type: "text", text: `Results for: ${query}` }],
    };
  }
  throw new Error("Tool not found");
});


const transport = new StdioServerTransport();
await server.connect(transport);

```

--------------------------------

### Verify Cosine CLI Installation

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

After installation via the one-line script, open a new terminal or source the output of the installer, then verify the installation by checking the CLI version.

```bash
cos --version
```

--------------------------------

### Build and Deploy Static Sites

Source: https://cosine.sh/docs/customizing/instant-sites

Build your site first, then deploy the generated static output directory. This is suitable for plain HTML/CSS/JS and static builds from frameworks.

```bash
pnpm build
cos deploy dist
```

--------------------------------

### Install CLI with Homebrew (macOS/Linux)

Source: https://cosine.sh/docs/cli/quickstart

Use this command to install the Cosine CLI on macOS or Linux using Homebrew. Verify the installation by checking the version.

```bash
brew install CosineAI/tap/cos
cos --version
```

--------------------------------

### Initialize Content Agent Folder

Source: https://cosine.sh/docs/learn/automating-content-research

Create a new directory for your blog posts and initialize it as a Cosine agent folder.

```bash
mkdir ~/blog-posts
cd ~/blog-posts
cos init
```

--------------------------------

### Example Workflow: Line Up Multiple Follow-up Tasks (Queue)

Source: https://cosine.sh/docs/cli/queueing-and-nudging

Steps to line up multiple follow-up tasks using the queue behavior in Cosine CLI.

```text
1. Ask Cosine to implement the main feature
2. While it is running, type: "Add tests for the new behavior" and press Ctrl+Q
3. Type: "Update the docs for this change" and press Ctrl+Q
```

--------------------------------

### Prevent Path Modification during Install

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Use the --no-modify-path flag with the one-line installation script to prevent the installer from automatically updating your shell startup files.

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --no-modify-path
```

--------------------------------

### Install Specific Cosine CLI Version

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Customize the one-line installation script to install a specific version of the Cosine CLI, such as a nightly build or a tagged release.

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --version nightly
```

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --version nightly-838
```

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --version v1.2.3
```

--------------------------------

### Show Production Configuration

Source: https://cosine.sh/docs/customizing/instant-sites

Shows the project’s current Instant Sites production configuration.

```bash
cos deploy config
```

--------------------------------

### Install Cosine CLI via Scoop (Windows)

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Add the CosineAI Scoop bucket and install the Cosine CLI using the Scoop package manager. This is a community-maintained installation method.

```powershell
scoop bucket add cosine https://github.com/CosineAI/scoop-bucket
scoop install cos
```

--------------------------------

### Deploy and Promote Site from CLI

Source: https://cosine.sh/docs/cloud/instant-sites-domains

This command deploys the built site and immediately promotes it to production in a single step. This is useful for ensuring the latest deployment is live.

```bash
cos deploy --promote dist
```

--------------------------------

### Install Cosine CLI to Specific Directory

Source: https://cosine.sh/docs/cli/install-and-authenticate-the-cli

Use the one-line installation script to install the Cosine CLI into a custom directory. This can be done using the --install-dir flag or the COSINE_INSTALL_DIR environment variable.

```bash
curl -fsSL https://cosine.sh/install | bash -s -- --install-dir "$HOME/bin"
```

```bash
COSINE_INSTALL_DIR="$HOME/bin" curl -fsSL https://cosine.sh/install | bash
```

--------------------------------

### Create and Navigate to a New Project Folder

Source: https://cosine.sh/docs/learn/managing-context-with-folders

Use these commands to create a new directory for your agent project and then change into that directory.

```bash
mkdir ~/social-posts
cd ~/social-posts
```

--------------------------------

### Example: Taking Screenshots with Browser Tools

Source: https://cosine.sh/docs/customizing/browser

Demonstrates how Cosine CLI uses `browser_navigate` and `browser_screenshot` tools to take a screenshot of a deployed application.

```text
You: Take a screenshot of my deployed app at https://myapp.com
Cosine: I'll navigate to your app and capture a screenshot.
[Uses browser_navigate and browser_screenshot tools]

```

--------------------------------

### Initialize Sales Outreach Folder

Source: https://cosine.sh/docs/learn/creating-a-sales-outreach-agent

Create a dedicated directory for your sales outreach agent and initialize the Cosine CLI.

```bash
mkdir ~/sales-outreach
cd ~/sales-outreach
cos init
```

--------------------------------

### Install Skill from GitHub Tree URL or Subpath

Source: https://cosine.sh/docs/customizing/skills

Installs a skill from a specific branch or subdirectory within a GitHub repository.

```bash
# Install from a GitHub tree URL or subpath
cos skills add https://github.com/anthropics/skills/tree/main/skills --skill skill-creator
```

--------------------------------

### Verify Claude Code CLI Installation

Source: https://cosine.sh/docs/cli/login-with-claude

After installing the Claude Code CLI, confirm that the 'claude' command is available in your terminal.

```bash
claude --version
```

--------------------------------

### Initialize a Summarization Agent Folder

Source: https://cosine.sh/docs/learn/building-a-summarization-agent

Create a new directory for your summarization agent and initialize it with the Cosine CLI. This sets up the necessary configuration files.

```bash
mkdir ~/call-summaries
cd ~/call-summaries
cos init
```

--------------------------------

### Deploy a Site from a Specific Path

Source: https://cosine.sh/docs/customizing/instant-sites

Deploy a website by providing the full path to its built output folder.

```bash
cos deploy ./apps/website/out
```

--------------------------------

### Example DNS Target

Source: https://cosine.sh/docs/cloud/instant-sites-domains

This is an example of a DNS target that a custom hostname should resolve to. Ensure your custom hostname resolves to this target.

```dns
cosine.page
```

--------------------------------

### Monorepo Setup: Mark Topmost Config

Source: https://cosine.sh/docs/cli/config/toml-config

Configure a TOML file to stop parent directory searches in monorepo setups.

```toml
is_topmost_config = true
```

--------------------------------

### Initialize Cosine Configuration

Source: https://cosine.sh/docs/cli/quickstart

Run this command to create repo-local configuration files for Cosine within your project directory.

```bash
cos init
```

--------------------------------

### Start Cosine CLI in Specific Directory

Source: https://cosine.sh/docs/cli/terminal-commands

Specify a working directory to start the Cosine TUI or run a one-shot task.

```bash
cos start --cwd /path/to/project
```

--------------------------------

### Example Production Subdomain

Source: https://cosine.sh/docs/cloud/instant-sites-domains

This is an example of a production subdomain that Cosine will use for your Instant Site. It forms the basis of the built-in production URL.

```text
docs-team.cosine.page
```

--------------------------------

### Example MCP Tool Usage

Source: https://cosine.sh/docs/mcp

Illustrates a user request and Cosine's response, indicating the use of an mcp_filesystem tool to list files.

```text
You: List all files in my projects directory
Cosine: I'll list the files for you. [Uses mcp_filesystem tool]
```

--------------------------------

### Database Query Example

Source: https://cosine.sh/docs/mcp/server-management

Example interaction where the user asks about tables in a database, and Cosine CLI uses the `mcp_postgres` tool to query the schema.

```text
You: What tables are in my postgres database?
Cosine: Let me check the database schema. [Uses mcp_postgres tool]
```

--------------------------------

### Interactive Project Setup Prompt

Source: https://cosine.sh/docs/customizing/instant-sites

This prompt appears when `cos deploy` is run in a folder without a configured Cosine project, offering to run `cos project init`.

```bash
No Cosine project is configured. Run `cos project init` now? [Y/n]:

```

--------------------------------

### Start a Session with Balanced Reasoning

Source: https://cosine.sh/docs/customizing/reasoning

Begin a session with a medium reasoning level, providing a good balance between speed and thoroughness.

```bash
# Balanced default
cos start --reasoning medium
```

--------------------------------

### Set Startup Mode with --mode

Source: https://cosine.sh/docs/cli/config/flags

Use the `--mode` flag to set the startup mode for Cosine. The selected mode persists across sessions. Accepted values include `normal`, `auto`, `plan`, and `swarm`.

```bash
# Start in plan mode
cos start --mode plan
```

```bash
# Start in auto mode
cos start --mode auto
```

```bash
# Start in swarm mode
cos start --mode swarm
```

```bash
# Start in manual mode (default)
cos start --mode manual
```

--------------------------------

### GitHub Operation Example

Source: https://cosine.sh/docs/mcp/server-management

Example interaction where the user requests to create a GitHub issue, and Cosine CLI uses the `mcp_github` tool to perform the action.

```text
You: Create an issue titled "Bug in login flow" in my repo
Cosine: I'll create that issue. [Uses mcp_github tool]
```

--------------------------------

### Start Agent with Mode and Reasoning Flags

Source: https://cosine.sh/docs/cli/changelog

Set the agent's mode and reasoning level directly at startup using `--mode` and `--reasoning` flags. Shorthand flags are available for common modes.

```bash
cos2 start --mode plan --reasoning high

```

```bash
cos2 start --am              # Start in auto mode

```

```bash
cos2 start --pm --reasoning high  # Plan mode with high reasoning

```

--------------------------------

### Deploy Site from CLI

Source: https://cosine.sh/docs/cloud/instant-sites-domains

Use this command to deploy the built site from your command-line interface. Ensure your build output is in the 'dist' directory.

```bash
cos deploy dist
```