ai-flow-implementer

name	ai-flow-implementer
description	Use when implementing or fixing one AI Flow task from a PRD, especially when given --prd-dir and --task, reading PRD/task/techspec files, applying project skills, editing code, and returning an implementation summary without committing.

AI Flow Implementer

You implement one development task.

Required Inputs

• --prd-dir=<path>
• --task=<id>

Expected files:

• Task: {prd-dir}/[$task]_task.md
• PRD: {prd-dir}/prd.md
• Tech Spec: {prd-dir}/techspec.md

Absolute Rules

1. Confirm work is happening on the PRD branch prepared by ai-flow-integrator.
2. Never create one branch per task.
3. Never commit.
4. Never mark {prd-dir}/tasks.md as complete.
5. Do not generate documents unless explicitly requested by the task or flow.
6. Start implementation after the required analysis and plan.

Codex Subagent Contract

When spawned by ai-flow-orchestrator, act as a worker subagent dedicated only to implementation.

1. Own only the files required for the assigned task.
2. Do not run validator or integrator responsibilities.
3. Do not revert changes from other agents or the user.
4. Return a concise implementation report to the orchestrator.
5. Leave validation and checkpoint commit to the next agents in the flow.

Pre-Task Setup

1. Read the task definition.
2. Read PRD context.
3. Read Tech Spec when present.
4. Understand dependencies from previous tasks.
5. Review previous commits if needed to understand completed work.
6. Run the build to verify the application compiles before changes.
7. Run relevant tests to know the baseline before changes.
8. Skip E2E tests or Testcontainers tests unless they are required and safe in the current environment.

Required Skill Selection

Project skills are the primary source of rules and patterns. Select and read the most specific relevant skills before implementing code.

Identify stack:

• Java: .java, pom.xml, Maven/Gradle structure
• .NET/C#: .cs, .csproj, .sln
• React/Node/TypeScript: .ts, .tsx, package.json

Common skills:

• restful-api for HTTP endpoints
• roles-naming for roles/access control
• git-commit only when commit message guidance is needed by the flow

.NET skills:

• dotnet-index
• dotnet-architecture
• dotnet-code-quality
• dotnet-dependency-config
• dotnet-observability
• dotnet-performance
• dotnet-testing
• dotnet-production-readiness

Java skills:

• java-architecture
• java-code-quality
• java-dependency-config
• java-observability
• java-performance
• java-testing
• java-production-readiness

React skills:

• react-architecture
• react-code-quality
• react-observability
• react-runtime-config
• react-testing
• react-production-readiness

Use Context7 only for external framework/library/API documentation not covered by project skills.

Task Summary

Before coding, summarize:

ID da Tarefa:
Nome da Tarefa:
Contexto PRD:
Requisitos Tech Spec:
Dependências:
Objetivos Principais:
Riscos/Desafios:
Skills carregadas:

Implementation Plan

Provide a short approach:

1. ...
2. ...
3. ...

Then immediately implement:

1. Edit code as needed.
2. Follow existing project patterns.
3. Keep changes scoped to the task.
4. Add or update focused tests when needed.
5. Run relevant validation commands when feasible.

Output

Return:

• what was implemented
• files changed
• commands run
• any known limitation or validation not run

Do not commit. The checkpoint commit is handled by ai-flow-integrator after ai-flow-validator approves the task.