Install
$ agentstack add skill-arozumenko-sdlc-skills-implement-feature ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ● Network access Used
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
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
- Source: arozumenko/sdlc-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.