implement

name	implement
description	End-to-end feature implementation from requirements to tests to documentation. Takes requirements file, GitHub issue, or description as input.
allowed-tools	Agent, Read, Write, Edit, Grep, Glob, Bash, Skill
user-invocable	true
argument-hint	<requirements-file|issue-number|'description'>
automation	autonomous
metadata	{"version":"1.0","created":"2026-03-13T00:00:00.000Z","author":"Ability.ai","changelog":["1.0: Initial version — end-to-end feature implementation from a requirements file, GitHub issue, or description through tests to documentation."]}

Implement Feature

ℹ️ First, set expectations: before anything else, print one short line with this skill's version and its most recent change — the top entry of metadata.changelog above — e.g. implement vX.Y — recent: <summary>. Then proceed.

End-to-end feature implementation from requirements to tested, documented code.

Purpose

Autonomously implement features by:

1. Parsing requirements from file, GitHub issue, or inline description
2. Understanding existing patterns via feature flows
3. Implementing with minimal necessary changes
4. Creating and running tests
5. Updating documentation

State Dependencies

Source	Location	Read	Write	Description
Requirements	docs/requirements/*.md	✅		Feature specs
GitHub Issues	GitHub repo	✅		Issue details
Feature Flows	docs/memory/feature-flows/	✅		Existing patterns
Architecture	docs/memory/architecture.md	✅		System design
Source Code	src/	✅	✅	Implementation
Tests	tests/	✅	✅	Test files

Arguments

• $ARGUMENTS:
  • File path: docs/requirements/MY_FEATURE.md
  • Issue number: #42 or 42
  • Inline description: "Add endpoint for X that does Y"

Process

Step 1: Parse Requirements

If file path provided: Read and parse the requirements document.

If issue number provided:

gh issue view ${ARGUMENTS#\#} --json title,body,labels

If inline description: Use the provided description directly as requirements.

Extract:

• Feature name/ID
• Core requirements (what must be built)
• Acceptance criteria (how to verify it works)
• Related features/flows (what to study first)

Step 2: Study Existing Patterns

Read relevant feature flows to understand existing patterns. Based on requirements, identify 2-4 related flows and read them.

Also read:

• docs/memory/architecture.md - for system context
• CLAUDE.md - for development rules

Step 3: Plan Implementation

1. Files to modify - list specific files with line ranges
2. Files to create - new files needed
3. Order of changes - dependencies between changes
4. Test strategy - what tests to write

Step 4: Implement Feature

Backend first (if applicable):

1. Database schema/migrations (if needed)
2. Service layer logic
3. Router/controller endpoints
4. Models/schemas

Frontend second (if applicable):

1. Store/state management
2. Components
3. Views/routing

Step 5: Create Tests

Coverage targets:

• Happy path (success case)
• Authentication/authorization
• Input validation
• Error cases
• Edge cases from requirements

Step 6: Run Tests

Invoke the test-runner agent to execute tests.

Step 7: Fix Test Failures

If tests fail:

1. Analyze failure
2. Identify root cause
3. Fix the issue
4. Re-run tests

Max iterations: 3 fix attempts.

Step 8: Sync Feature Flows

Invoke sync-feature-flows to update documentation:

/sync-feature-flows recent

Step 9: Update Documentation

Invoke update-docs to finalize:

/update-docs

Step 10: Report Completion

Output final status with changes made, tests results, and documentation updated.

Completion Checklist

• Requirements parsed from input
• Related feature flows studied
• Implementation plan created
• Feature implemented
• Tests created
• Tests run via test-runner agent
• All tests passing (or max fix attempts reached)
• Feature flows synced
• Documentation updated
• GitHub issue updated (if issue number was provided)
• Completion report generated

Error Recovery

Error	Recovery
Requirements file not found	Ask for correct path or use inline description
GitHub issue not found	Verify issue number, check repo access
Test runner unavailable	Run tests directly
Tests fail after 3 fix attempts	Report current state, list remaining failures
Feature flow sync fails	Create flow manually using feature-flow-analysis

Best Practices Enforced

• Minimal necessary changes - only modify what's required
• No unsolicited refactoring - don't "improve" unrelated code
• Follow existing patterns - match code style
• No over-engineering - simplest solution that works