# Implement Feature

> Use when the user says "implement", "build feature", "work on task", or once a plan or story is approved and you're about to start building. The end-to-end feature-implementation workflow; delegates the final handoff to completing-a-task.

- **Type:** Skill
- **Install:** `agentstack add skill-arozumenko-sdlc-skills-implement-feature`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [arozumenko](https://agentstack.voostack.com/s/arozumenko)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [arozumenko](https://github.com/arozumenko)
- **Source:** https://github.com/arozumenko/sdlc-skills/tree/main/bundles/feature-development/skills/implement-feature

## Install

```sh
agentstack add skill-arozumenko-sdlc-skills-implement-feature
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

## Feature Implementation: Plan → Test → Build → Verify → Ship

**Core philosophy:** understand the plan, write tests first, implement to pass
them, verify everything, ship clean.

## Platform & systems

`gh issue …` / `gh pr …` below are the **GitHub reference**. Translate to your
tracker per `.agents/profile.md` § Project systems (jira / gitlab-issues /
azure-boards / linear) and your code host's CLI per `.agents/workflow.md`
(GitLab `glab`, Azure DevOps `az repos`, Bitbucket `bb`, Gitea `tea`). If scout
hasn't recorded them, ask before assuming GitHub.

## The eight steps

```
1. Load the plan
2. Write test cases
3. Implement
4. Manual verification
5. Write automated tests
6. Run full test suite
7. Commit & PR
8. Document on ticket
```

### 1. Load the plan

`gh issue view `. Identify acceptance criteria (definition of done),
interface contract (input/output types if defined), dependencies, and what's
explicitly in and out of scope. If anything is unclear, ask the tech-lead via
a subagent call or tracker comment *before* writing code. Post a `🔧 **Started**` comment so the team
knows work is in progress.

### 2. Write test cases (TDD red phase)

Define "done" as executable tests before implementing. For each acceptance
criterion, write a test. Cover happy path, error cases, edge cases. Run them
and confirm they fail — that's the red phase.

### 3. Implement

Write the minimum code to pass the tests. Rules:

- **Follow existing patterns** — read similar features in the codebase first
- **One change at a time** — verify syntax after each file edit
- **Match the interface contract exactly** if the task defines one
- **No scope creep** — don't add features beyond the acceptance criteria
- **Don't refactor neighbors** — a feature task is not a cleanup task
- **Don't optimize prematurely** — make it work before making it fast

### 4. Manual verification

Sanity check before running automated tests. API → `curl` the endpoint. UI →
browser snapshot + interaction. Backend logic → one-liner that calls the
function. Catches the obvious breaks before test run.

### 5. Write automated tests

Extend the test cases from step 2 with integration and E2E tests if the
feature touches external systems (DB, APIs) or has a UI component.

### 6. Run full test suite

All tests, lint, type check, and `git diff --stat` to confirm the scope of
changes matches expectations. Don't ship if anything fails — fix it first.

### 7. Commit & PR

Feature branch, focused commit, `gh pr create` with proper title and body.
The `completing-a-task` skill has the full protocol.

### 8. Document on ticket

`✅ **Done**` comment with summary, test counts, PR number, and key
architectural decisions. Report in your final reply that the PR is ready for
review.

## Command recipes

All the test templates, verification commands, PR body heredocs, and handoff
notification snippets live in [references/commands.md](references/commands.md). Load
that file when you need exact command syntax.

## Anti-Patterns

- Don't implement without understanding the acceptance criteria
- Don't skip tests — they're proof of correctness, not overhead
- Don't expand scope beyond what the task defines
- Don't commit without running the full test suite
- Don't forget to document on the ticket — the audit trail matters

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [arozumenko](https://github.com/arozumenko)
- **Source:** [arozumenko/sdlc-skills](https://github.com/arozumenko/sdlc-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** yes
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-arozumenko-sdlc-skills-implement-feature
- Seller: https://agentstack.voostack.com/s/arozumenko
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.