Your first AI flow

In this guide you'll complete the full AI-assisted work cycle: create a task in Almirant, implement it with Claude Code using the /implement skill, and review it with /review-task. By the end, you'll have an implemented and validated task without manually writing a single line of code.

Full flow

Requirements

Before you begin, make sure you have:

• A project created in Almirant with at least one board -- Your first board
• Claude Code connected via MCP -- Connect with Claude Code
• A code repository where the AI can work

Step 1: Create a task in Almirant

From your board in Almirant, create a new work item of type Task:

Field	Example value
Title	"Add GET /api/health endpoint"
Type	Task
Description	"Create a health check endpoint that returns status 200 with timestamp and API version"
Priority	Medium

Once created, the work item will receive an automatic identifier (for example, MC-T-1). Note this identifier -- you'll need it in the next step.

Good descriptions = better implementation

The more detailed the work item description, the better the AI implementation will be. Include acceptance criteria, expected response examples, and any relevant constraints.

Step 2: Implement with /implement

Open Claude Code in your terminal, inside the project repository. Use the /implement skill followed by the task identifier:

> /implement MC-T-1

The AI will automatically execute the following steps:

1. Reads the task -- Gets the title, description, and context of the work item from Almirant
2. Analyzes the codebase -- Understands the project structure and existing conventions
3. Implements -- Writes the code needed to complete the task
4. Moves the item -- Updates the work item status to "Review" on the board

Automatic movement

When running /implement, the AI automatically moves the work item to the "In Progress" column while working, and to "Review" when the implementation is finished.

Step 3: Review the changes

Once the AI finishes the implementation, you can review the generated changes. Use git diff or your preferred review tool to verify the code.

For an AI-assisted review, use the /review-task skill:

> /review-task MC-T-1

The AI will review its own implementation (or that of any other contributor) looking for:

• Logic or syntax errors
• Compliance with the task description
• Adherence to project conventions
• Potential performance or security issues

Step 4: Complete the cycle

After the review, you have several options:

Action	Skill	Result
Validate that the task meets the criteria	/validate MC-T-1	Verifies acceptance criteria
Run tests	/test-task MC-T-1	Creates and runs tests for the task
Create Pull Request	/pr MC-T-1	Generates a PR with the changes

When you're satisfied with the implementation, move the work item to the Done column on your board, or let the AI do it upon completing the validation.

Available skills

These are the skills you can use in your workflow:

Skill	Description	Example
/implement	Implements the code for a task	/implement MC-T-1
/review-task	Reviews the code for a task	/review-task MC-T-1
/validate	Validates that the implementation meets the criteria	/validate MC-T-1
/test-task	Generates and runs tests for the task	/test-task MC-T-1
/pr	Creates a Pull Request with the changes	/pr MC-T-1
/ideate	Generates ideas and subtasks from an objective	/ideate "improve dashboard performance"

Complete example

Here's a typical complete session in Claude Code:

# 1. Implement the task
> /implement MC-T-1

# Claude reads the task, analyzes the codebase, and implements...
# Work item moved to "Review"

# 2. Review the implementation
> /review-task MC-T-1

# Claude reviews the code and suggests improvements if needed...

# 3. Run tests
> /test-task MC-T-1

# Claude generates tests and runs them...

# 4. Create Pull Request
> /pr MC-T-1

# Claude creates a PR with title, description, and reference to the work item

Always review the changes

Although the AI is very capable, always review the changes before merging. The AI is an assistance tool, not a substitute for human review.

What's next

You've mastered the basic AI-assisted workflow in Almirant. From here you can explore:

• Boards and columns -- Customize your workflow with columns and advanced settings
• Work Items -- Learn about the full hierarchy (Epic > Feature > Story > Task)
• Sprints -- Plan your work in iterations
• GitHub integrations -- Connect your repository and automate the workflow