How to use the D365 Solution Blueprint agent skill

What the d365-solution-blueprint skill makes the agent do

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches the skill's description. This one is for authoring a Dynamics 365 Finance and Supply Chain Management Solution Blueprint from scratch. The agent acts as the solution architect and works through a structured, section-by-section interview rather than generating a finished document in one go.

The blueprint covers scope, target operating model, application and data architecture, integration landscape, migration strategy, security model, ALM, testing, deployment, and support approach. Each architectural decision is recorded in a decision log with its rationale and the alternatives that were rejected. Anything not yet settled is marked as open rather than filled in with an invented answer.

• Asks focused questions, usually three to five at a time
• Offers options with trade-offs where a real architectural choice exists, and gives a recommendation
• Records each decision with rationale, rejected alternatives, implications, decider and date
• Flags eight load-bearing decisions, such as legal entity structure and dual-write scope, so they are not skipped
• Saves a working blueprint file between sessions so you can resume later

Install the skill from the ZIP download

Download the skill as a ZIP file from the page where you found this guide. Then choose one of the two installation routes below.

For a personal install, unzip the file into your user skills folder at ~/.claude/skills/. To limit the skill to one project, unzip it into the .claude/skills/ folder inside that project instead.

If you use the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills instead of unzipping it yourself.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Trigger the skill with a natural request

You do not need to type a special command. The agent matches your request against the skill's description, which lists the situations the skill is for. Requests that clearly describe creating D365 architecture documentation are the most reliable triggers.

The skill is also built to stay out of the way for review work. If you ask it to critique an existing design, it is designed not to apply, because reviewing a design is a different task from authoring a blueprint.

• "Help me prepare a Solution Blueprint for our D365 Finance and SCM implementation."
• "We are starting a Dynamics 365 implementation. Walk me through the architecture decisions we need to make."
• "Design the architecture for our new D365 programme and record the decisions as we go."
• "Identify the architectural decisions our programme must make before build begins."

What a blueprint session looks like

The work is organised into five tracks. Track A (Foundation) must come first, because legal entity structure and phasing decisions affect every later section. Tracks B to E can follow in the order you prefer. Each section follows the same five steps: frame it, ask questions, propose options, record the decision or open item, then draft and save.

The agent will not run two sections in one turn unless you explicitly ask it to move faster. This is deliberate. The value comes from answering the questions carefully.

• Track A: programme context, scope, target operating model
• Track B: application, data, and integration architecture
• Track C: data migration, security, compliance, and licensing
• Track D: environment and ALM, reporting, performance and volumetrics
• Track E: test strategy, deployment and cutover, support and operating model

Resume a blueprint in a later session

The working blueprint file is the record between sessions. At the end of each session the agent saves or updates it in the available workspace and tells you which file holds the current state. Its progress tracker shows each section's status and open items.

When you come back, start by pointing the agent at that file. The skill tells it to read the progress tracker and decision log first and to summarise open items before continuing. If there is no persistent workspace copy, the agent will ask you for the latest file rather than reconstruct decisions from memory.

• Keep the saved blueprint file and share it when you resume
• Expect the agent to summarise open items before new questions
• Expect it not to re-ask questions the decision log already answers

What you need and what the skill does not do

The skill works best in a workspace where the agent can save and update a working blueprint file between sessions. Without that, every session starts from whatever you paste in. The skill runs no scripts and makes no network calls.

It covers blueprint-level decisions only. When a discussion moves into detailed interface specifications, role catalogues, timed cutover runbooks, or formal project health reviews, the skill either hands off to a specialist skill if one is installed or keeps the blueprint at architecture-decision depth and names the follow-on deliverable.

• Requires: a workspace where the agent can save and update the blueprint file (recommended)
• Runs scripts: no
• Uses network: no
• Included files: SKILL.md, assets/blueprint-template.md, references/section-guide.md

Safety notes and limits

The main risk is not the skill's code, which runs nothing, but over-trusting the output. The skill is designed to leave gaps open. A blueprint with unresolved items is more honest than one with invented answers, and the output should not be built from without expert validation and real client sign-off.

Treat every assumption in the document as something to verify. Assumptions are labelled in the text and should have an owner and a validation date.

• Review every entry marked as an assumption before relying on it
• Treat open items as blocking until someone with authority decides them
• Do not present the draft to a client as final without review by an experienced architect

Tips for getting useful output

Bring your actual programme facts to the first session, including number of legal entities, geographies, and planned phasing. The skill asks for these early because Track A decisions cascade through the rest of the blueprint.

Answer with the decision owner's name where you know it. The decision log requires a named person, not "the project", and the skill will ask for one.

• Start with Track A, even if you want to focus on integration first
• Keep answers short and specific; the skill asks three to five questions at a time
• If you cannot decide a load-bearing item, ask the agent to record it as open with an owner and a needed-by date