Installation

Prerequisites

Before you begin, make sure you have:

• VS Code (latest version recommended)
• GitHub Copilot subscription with agent mode enabled
• GitHub Copilot Chat extension installed and signed in
• Agent mode enabled in VS Code settings ("github.copilot.chat.agent.enabled": true)

---

Option A — Copy into your project (recommended)

This is the simplest approach. The kit travels with your repository.

1. Copy the .github/agents/ folder to your project's .github/agents/
2. Copy the .github/prompts/ folder to your project's .github/prompts/
3. Copy the .github/templates/ folder to your project's .github/templates/
4. Add the following to your project's .vscode/settings.json:

{
  "chat.agentFilesLocations": { ".github/agents": true },
  "chat.promptFilesLocations": { ".github/prompts": true }
}

5. Create the doc-specs/ folder at the root of your project:

mkdir doc-specs

---

Option B — Point VS Code to the kit folder or an unpacked bundle

If you prefer to keep the kit as a shared external folder (e.g., used across multiple projects), add the paths to your .vscode/settings.json:

{
  "chat.agentFilesLocations": { "./ai-sdlc-kit/.github/agents": true },
  "chat.promptFilesLocations": { "./ai-sdlc-kit/.github/prompts": true }
}

Bash-first installation (Option B)

To avoid editing settings.json manually, run the installer from the consumer project root so it updates that project's .vscode/settings.json:

bash /path/to/ai-sdlc-kit/scripts/install.sh /path/to/ai-sdlc-kit

For example:

cd /path/to/your-project
bash /path/to/ai-sdlc-kit/scripts/install.sh /path/to/ai-sdlc-kit

If you copied the kit into the consumer repo as ./ai-sdlc-kit, you can still run:

cd /path/to/your-project
bash ./ai-sdlc-kit/scripts/install.sh ./ai-sdlc-kit

The script detects an existing settings.json and merges the required entries without overwriting any other configuration. If the file does not exist, it is created.

---

Offline and governed environments

You do not need a second documentation source to run the kit. The operational guidance now lives directly in the official docs under:

• Operational Modes
• Governed Environments

To generate an approved offline artifact for internal distribution:

bash scripts/package-bundle.sh

This produces a versioned bundle directory, a .tar.gz archive, and a sha256 file under dist/.

---

Verify discovery

After installation, open the GitHub Copilot Chat panel and type / in the input field. You should see the kit prompts listed (e.g., /discovery-refine, /task-implement, etc.).

Open the agent selector and confirm that the kit agents appear (e.g., 🧭 Discovery, 🏗️ Architect, 🛠️ Implementer, etc.).

If prompts or agents are not discovered:

1. Confirm the paths in .vscode/settings.json are correct and relative to the workspace root.
2. Reload VS Code (Ctrl+Shift+P → Developer: Reload Window).
3. Check that the files in .github/agents/ have the .agent.md extension and valid YAML frontmatter.

---

Project folder structure after installation

your-project/
├── .github/
│   ├── agents/          ← AI SDLC Kit agents
│   ├── prompts/         ← AI SDLC Kit prompts
│   └── templates/       ← AI SDLC Kit templates
├── .vscode/
│   └── settings.json    ← agent/prompt discovery config
├── doc-specs/           ← all spec artefacts go here
│   └── idea.txt         ← (Flow A) or
│   └── task.txt         ← (Flow B)
└── ...

The doc-specs/ folder is the working directory for all generated artefacts — specs, PRDs, epics, decisions logs, and context memory. It should be committed to version control.