Feature Workflow Skill

Status: Active Last Updated: 2026-01-08

Overview

This skill guides the complete feature development lifecycle from idea to deployment.

Quick Reference

Work Item Routing

I have...	Put it in...	Table
An idea	Feedback	`claude.feedback` (type='idea')
A bug	Feedback	`claude.feedback` (type='bug')
A question	Feedback	`claude.feedback` (type='question')
A feature to build	Features	`claude.features`
A task to do	Build Tasks	`claude.build_tasks`
Work right now	TodoWrite	Session only

Feature Lifecycle

code

IDEA → FEEDBACK → FEATURE → TASKS → IMPLEMENTATION → REVIEW → DEPLOY

1. Capture Idea

sql

INSERT INTO claude.feedback (
    feedback_id, project_id, feedback_type, description, priority, status
) VALUES (
    gen_random_uuid(),
    'project-uuid-here',
    'idea',  -- or 'design', 'question', 'change'
    'Description of the idea',
    3,  -- 1=critical, 5=backlog
    'new'
);

2. Promote to Feature

When an idea is approved for implementation:

sql

-- Create feature
INSERT INTO claude.features (
    feature_id, project_id, feature_name, description, priority, status
) VALUES (
    gen_random_uuid(),
    'project-uuid-here',
    'Feature Name',
    'Detailed description',
    2,
    'planned'
);

-- Link and close feedback
UPDATE claude.feedback 
SET status = 'implemented', 
    resolved_at = NOW()
WHERE feedback_id = 'feedback-uuid';

3. Break Into Tasks

sql

INSERT INTO claude.build_tasks (
    task_id, feature_id, task_name, description, status
) VALUES 
    (gen_random_uuid(), 'feature-uuid', 'Create API endpoint', '...', 'pending'),
    (gen_random_uuid(), 'feature-uuid', 'Add database table', '...', 'pending'),
    (gen_random_uuid(), 'feature-uuid', 'Write unit tests', '...', 'pending');

4. Track Progress

Use TodoWrite for session-level tracking:

json

[
  {"content": "Create API endpoint", "status": "in_progress", "activeForm": "Creating API endpoint..."},
  {"content": "Add database table", "status": "pending", "activeForm": "Adding database table..."},
  {"content": "Write unit tests", "status": "pending", "activeForm": "Writing unit tests..."}
]

Status Values

Feedback Status

•new - Just created
•in_progress - Being worked on
•fixed - Issue resolved
•wont_fix - Won't be addressed

Feature Status

•planned - Approved for development
•in_progress - Under active development
•completed - Fully implemented
•cancelled - Not going forward

Task Status

•pending - Not started
•in_progress - Active work
•completed - Done
•blocked - Waiting on dependency

Priority Scale

Priority	Meaning	SLA
1	Critical	Immediate
2	High	This sprint
3	Medium	This quarter
4	Low	When time permits
5	Backlog	Maybe someday

Common Queries

sql

-- Open features for a project
SELECT feature_name, status, priority 
FROM claude.features 
WHERE project_id = 'project-uuid' AND status != 'completed'
ORDER BY priority;

-- Tasks for a feature
SELECT task_name, status 
FROM claude.build_tasks 
WHERE feature_id = 'feature-uuid'
ORDER BY created_at;

-- All open work items
SELECT 'feedback' as source, description, priority 
FROM claude.feedback WHERE status IN ('new', 'in_progress')
UNION ALL
SELECT 'feature', feature_name, priority 
FROM claude.features WHERE status IN ('planned', 'in_progress')
ORDER BY priority;

Key Patterns

1. Auto-TodoWrite for Workflows

When process-guidance is injected, immediately add todos:

json

[
  {"content": "[PROC-ID] Step 1", "status": "pending", "activeForm": "Doing step 1..."},
  {"content": "[PROC-ID] Step 2", "status": "pending", "activeForm": "Doing step 2..."}
]

2. Feature Documentation

Create docs alongside features:

markdown

# Feature: {Feature Name}

## Problem
What problem does this solve?

## Solution
How does it solve it?

## Implementation
- Step 1
- Step 2

## Testing
How to verify it works?

Related Skills

•database-operations - Data storage patterns
•testing-patterns - Test before deploy
•code-review - Review process

Version: 1.0