Skip to main content

Stage 6: Implementation

Overview

Implementation is the highest-density stage in the project lifecycle, with 30 assets spanning agents, prompts, instructions, and skills. This stage covers coding, content creation, prompt engineering, data analysis, and infrastructure work. The RPI (Research, Plan, Implement) methodology provides structured execution guidance for complex tasks.

When You Enter This Stage

You enter Implementation after completing Stage 5: Sprint Planning with assigned work items. You also re-enter this stage from Stage 7: Review when rework is needed, from Stage 8: Delivery at the start of each new sprint, or from Stage 9: Operations for hotfixes.

NOTE

Prerequisites: Sprint planned with assigned work items. Development environment configured from Stage 1: Setup.

Available Tools

Primary Agents

ToolTypeHow to InvokePurpose
rpi-agentAgentSelect rpi-agent agentOrchestrate the full research-plan-implement workflow
task-researcherAgentSelect task-researcher agentResearch requirements and gather codebase evidence
task-plannerAgentSelect task-planner agentCreate implementation plans from research findings
task-implementorAgentSelect task-implementor agentBuild components following plans
task-reviewerAgentSelect task-reviewer agentValidate implementation against plan and research
gen-jupyter-notebookAgentSelect gen-jupyter-notebook agentCreate data analysis notebooks
gen-streamlit-dashboardAgentSelect gen-streamlit-dashboard agentGenerate Streamlit dashboards
prompt-builderAgentSelect prompt-builder agentCreate and refine prompt engineering artifacts

Supporting Agents

ToolTypeHow to InvokePurpose
phase-implementorAgentSelect phase-implementor agentExecute individual implementation phases
prompt-updaterAgentSelect prompt-updater agentUpdate existing prompts and instructions
researcher-subagentAgentSelect researcher-subagent agentConduct focused research within tasks

Prompts

ToolTypeHow to InvokePurpose
rpiPrompt/rpiStart the full RPI workflow
task-researchPrompt/task-researchResearch requirements for a task
task-planPrompt/task-planCreate an implementation plan from research
task-implementPrompt/task-implementBegin implementation of a specific task
task-reviewPrompt/task-reviewReview implementation against the plan
prompt-buildPrompt/prompt-buildCreate a new prompt engineering artifact
prompt-analyzePrompt/prompt-analyzeAnalyze prompt quality and effectiveness
prompt-refactorPrompt/prompt-refactorRefactor and improve existing prompts
git-commitPrompt/git-commitStage and commit changes
git-commit-messagePrompt/git-commit-messageGenerate a commit message for staged changes

Auto-Activated Instructions

All coding standard instructions activate automatically based on file type:

InstructionActivates OnPurpose
csharp**/*.csC# coding standards
python-script**/*.pyPython scripting standards
bash**/*.shBash script standards
bicep**/bicep/**Bicep infrastructure standards
terraform**/*.tfTerraform infrastructure standards
workflows.github/workflows/*.ymlGitHub Actions workflow standards
markdown**/*.mdMarkdown formatting rules
writing-style**/*.mdVoice and tone conventions
prompt-builderAI artifactsPrompt engineering authoring standards
hve-core-location**Reference resolution for hve-core

Skills

ToolTypeHow to InvokePurpose
video-to-gifSkillReferenced in chatConvert video to optimized GIFs

Role-Specific Guidance

Engineers are the primary users of Implementation, spending the majority of their engagement time here. Tech Leads contribute architecture-sensitive implementations. Data Scientists use notebook and dashboard generators. SREs handle infrastructure code. New Contributors start with guided tasks.

Starter Prompts

Full RPI Workflow

/rpi Implement the pagination logic for the /api/v2/search endpoint.
Add cursor-based pagination with a default page size of 50 and a maximum
of 200 results per request. Follow the existing pagination pattern in
src/api/handlers/list-resources.py.

Step-by-Step RPI Agents

Use individual task agents when you want more control over each phase.

/task-research Investigate how the existing list-resources handler in
src/api/handlers/list-resources.py implements pagination. Identify the
cursor encoding strategy, default and maximum page sizes, and response
envelope structure.

After research completes, plan the implementation:

/task-plan Create an implementation plan for adding cursor-based pagination
to the /api/v2/search endpoint following the patterns documented in the
research output.

Execute the plan:

Select task-implementor agent:

Build the webhook delivery system following the plan in
.copilot-tracking/plans/webhook-delivery-plan.md. Start with the event
dispatcher component and implement the retry queue second.

Select gen-jupyter-notebook agent:

Create a data analysis notebook for the Q4 sales transactions dataset in
data/sales-q4-2025.parquet. Include data quality assessment, revenue trend
analysis by product category and region, and customer cohort segmentation
using RFM scoring with matplotlib visualizations.

After implementation, validate the changes:

/task-review Validate the pagination implementation against the plan.
Check cursor encoding, page size limits, response envelope consistency,
and error handling for invalid cursor values.

Stage Outputs and Next Stage

Implementation produces source code, documentation, notebooks, dashboards, prompt artifacts, and infrastructure definitions. Transition to Stage 7: Review when implementation is complete. Use /clear to reset context before starting the review cycle.

🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers.