token-optimization

A customer-facing engagement for context management and token optimization

This repository contains source material for a practical developer workshop on context management, token optimization, agent customization, tool/MCP hygiene, model choice, AI evals, usage visibility, and sustainable team practices.

Start with labs/README.md for 1-hour, 2-hour, and 4-hour delivery outlines, then use the chapter files in order or as modular source material.

Sample web app

This repository also includes a static JavaScript sample app for GitHub Copilot Usage-Based Budgeting.

To run it locally:

python3 -m http.server 8080

Then open http://localhost:8080 in a browser.

Copilot primitives

This repo now includes a minimal set of Copilot customization assets for building workshop-ready token optimization content:

• .github/copilot-instructions.md for concise repository guidance
• .github/agents/workshop-content-builder.agent.md for GPT-5.5 content creation across labs, decks, and teaching materials
• .github/agents/cost-rubber-duck.agent.md for Claude Opus 4.7 second-opinion reviews
• .github/prompts/build-workshop-content.prompt.md for repeatable workshop-content generation
• .github/skills/source-material-ingestion/SKILL.md for .pptx and .csv ingestion workflows
• .github/skills/workshop-authoring/SKILL.md for labs, decks, and facilitator material workflows
• .vscode/mcp.json for Microsoft Learn and source-file ingestion via MarkItDown MCP

Why skills here

The repo uses small project skills for workflows that should only load when relevant. That keeps always-on instructions shorter and aligns with the token-optimization guidance taught by the workshop.

Tool setup notes

• The explicit file-ingestion tool is markitdown in .vscode/mcp.json.

• This is not a native built-in reader in the repo by itself. .vscode/mcp.json only declares the MCP server the client should use.

• Use MarkItDown to open and convert both .pptx and .csv files into markdown before reusing them in labs, prompts, or presentations.

• Install MarkItDown MCP with PPTX support before using the PowerPoint reader flow:

  pip install "markitdown[pptx]" markitdown-mcp

• After the MCP server is installed and your Copilot client loads workspace MCP servers, pass in a .pptx or .csv file by attaching the file or by asking the agent to read that file path with the markitdown tool.

• Use Microsoft Learn MCP for current Microsoft guidance.

• CSV support is included in MarkItDown's core text-based format handling.

• Treat PPTX conversion as full textual extraction for slides, tables, and reusable deck content, not full-fidelity inspection of animations, embedded media, or every binary object.

• Use the agent's web tools for current GitHub documentation when the answer depends on latest behavior.

Workshop goals

• Help customers reduce avoidable AI usage and improve answer quality.
• Build trust after billing and usage model changes by showing transparent, practical controls.
• Give attendees repeatable habits they can apply in other projects.
• Introduce eval-driven improvement so teams can measure whether changes help.

Contents

• labs/README.md — overview, prerequisites, and timed agendas
• labs/01-context-management-basics.md
• labs/02-instructions-and-agent-customizations.md
• labs/03-mcp-and-tool-optimization.md
• labs/04-chat-session-management.md
• labs/05-model-choice.md
• labs/06-chat-history-and-memory.md
• labs/07-usage-and-billing-visibility.md
• labs/08-ai-evals-and-observability.md
• labs/09-ideal-workshop-repo.md
• labs/10-next-steps-and-extra-topics.md