| name | feature-implementation |
| description | Use when implementing a feature from a completed spec in this .NET 10 Minimal API + Vue 3 project. Orchestrates architect, backend-developer, frontend-developer, test-writer, and code-reviewer with quality gates and fix loops. |
Feature Implementation Skill
Purpose
Use this skill after a feature has been documented and approved.
It coordinates subagents in a deterministic sequence so work is planned, implemented, verified, and reviewed before completion.
This skill is the canonical home for long-form feature workflow guidance, delivery gates, and fix-loop policy. Keep always-on and file-scoped instructions limited to routing and constraints.
Project Context
- Backend: .NET 10 Minimal API (C#), Clean Architecture + DDD (Core → Application → Infrastructure)
- Frontend: Vue 3 (Composition API) + Vite + Pinia + Tailwind CSS
- Data: MySQL 8.0+ with raw parameterized SQL (no ORM)
- Backend tests:
server/Mongoose.Api.Tests/ (xUnit, TestWebApplicationFactory)
- Frontend tests:
client/test/unit/ (Vitest + Vue Test Utils)
- E2E tests:
client/e2e/ (Playwright)
Key References
Inputs Required
- Feature spec path under
.github/specs/features/
- Definition of done (acceptance criteria)
- Scope constraints (in scope / out of scope)
Output
- Implemented feature aligned with spec
- Updated tests covering changed behavior
- Review findings addressed or explicitly tracked
- Final implementation summary with changed files
Recommended Orchestration Model
You do not need a separate orchestrator agent to start.
This skill itself should orchestrate the workflow and call specialized subagents.
Create a dedicated orchestrator agent only if one or more of these are true:
- You want strict tool restrictions per phase.
- You want reusable delivery governance across many repositories.
- You need compliance gates beyond normal review and testing.
Agent Sequence
Step 1: Architecture Breakdown
Agent: architect
Goals:
- Break feature into backend, frontend, data, and test work items.
- Identify risks, dependencies, migration needs, and non-functional constraints.
- Produce a phased implementation plan with acceptance checks.
- Ensure DDD alignment: bounded contexts, ubiquitous language, domain model placement.
Required deliverable:
- Task breakdown with ordered milestones and risk notes.
Contract freeze output (required before implementation):
- Interface contracts between streams (endpoint routes, request/response DTOs, shared data shapes).
- Validation and error-state contract.
- Work stream definitions: which agent handles each stream, and which files it owns.
Step 2: Implementation Streams
The architect defines one or more independent work streams — each assigned to the most appropriate agent. Streams run sequentially (subagent calls are blocking), but are designed to be independent so they do not conflict.
Common stream configurations:
- Backend + Frontend — one
backend-developer stream, one frontend-developer stream.
- Multiple frontend streams — e.g., two
frontend-developer invocations building different dashboard cards.
- Backend-only — a single
backend-developer stream when no UI changes are needed.
- Any mix — the architect decides based on the feature shape.
Stream preconditions (enforced for every stream):
- Architect contract freeze is complete.
- File ownership map has no overlap between streams.
- Any shared contracts (API shapes, store interfaces) are documented before streams begin.
Backend stream rules (when applicable)
Agent: backend-developer
Required deliverable: code changes plus test updates and implementation notes.
Frontend stream rules (when applicable)
Agent: frontend-developer
Required deliverable: code changes plus Vitest unit tests and UX notes.
Step 3: Sync Gate (Post-Implementation)
Owner: orchestrating skill execution
Goals:
- Reconcile integration points across all streams.
- Confirm contracts still match implementation.
- Resolve any cross-stream conflicts before verification.
Required checks:
- Endpoint routes and DTO shapes match contract.
- Payload and validation behavior match contract.
- Error and empty states are consistent across all streams.
Step 4: Verification
Agent: test-writer
Goals:
- Verify acceptance criteria against implemented behavior.
- Validate regression risk in touched areas.
- Confirm tests are deterministic and meaningful.
- Use the run-all-tests skill to execute the full suite (backend xUnit → frontend Vitest → Playwright E2E).
Required deliverable:
- Verification report with pass/fail criteria and defects.
Step 5: Code Review
Agent: code-reviewer
Goals:
- Review for correctness, maintainability, security, and test adequacy.
- Use
copilot-instructions.md and the relevant targeted instruction files as the review standard.
- Prioritize issues by severity and identify required fixes.
Required deliverable:
- Findings list ordered by severity, with file references.
Fix Loop Policy
If verification or review finds issues:
- Route each issue to the agent that owns the affected stream (per the file ownership map).
- Route cross-cutting design issues to architect.
- If multiple streams are impacted, run fixes sequentially with a mini contract freeze to prevent conflicts.
- Re-run sync gate checks.
- Re-run test-writer after fixes.
- Re-run code-reviewer for final sign-off.
Repeat until:
- Acceptance criteria are met.
- Tests pass (xUnit, Vitest, and Playwright).
- No critical review findings remain.
Guardrails
- Do not start implementation without a feature spec.
- Keep changes scoped to approved feature boundaries.
- Avoid unrelated refactors unless required for correctness.
- Keep commits/review units small and traceable.
- Follow repo-wide invariants from
copilot-instructions.md and file-scoped rules from the targeted instruction files.
Suggested Invocation Prompt
Use this skill when you say:
- Implement feature from spec at [path]
- Run full feature delivery workflow
- Plan, build, verify, and review this feature
Assumptions To Confirm At Start
- Spec is approved and stable enough to implement.
- Database migration strategy is agreed (if schema changes are needed).
- Expected rollout strategy is known (feature flag, direct release, etc.).