Conversation
Add tools for maximizing GitHub Copilot effectiveness through better context management: - Instructions: Guidelines for structuring code so Copilot understands it - Agent: Context Architect - plans multi-file changes by mapping dependencies - Prompts: - context-map: Map all affected files before changes - what-context-needed: Ask Copilot what files it needs - refactor-plan: Create phased refactor plans with rollback steps Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Adds a new Context Engineering collection to the repository, bundling Copilot prompts, an agent, and instructions aimed at improving context management for better Copilot outcomes.
Changes:
- Added three new prompt templates: Context Map, Refactor Plan, and What Context Do You Need?
- Added a new instruction file (Context Engineering) and a new agent (Context Architect)
- Added a new collection (Markdown + YAML manifest) and updated the generated docs indexes for prompts/instructions/agents/collections
Reviewed changes
Copilot reviewed 11 out of 11 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| prompts/what-context-needed.prompt.md | New prompt to ask which files Copilot needs before answering |
| prompts/refactor-plan.prompt.md | New prompt for phased multi-file refactor planning |
| prompts/context-map.prompt.md | New prompt to map relevant files/dependencies/tests before changes |
| instructions/context-engineering.instructions.md | New instruction set for structuring projects/code to improve Copilot context |
| agents/context-architect.agent.md | New agent for mapping dependencies and sequencing multi-file changes |
| collections/context-engineering.md | New collection landing page describing items + usage guidance |
| collections/context-engineering.collection.yml | New collection manifest defining included items and display metadata |
| docs/README.prompts.md | Adds the new prompts to the prompts index table |
| docs/README.instructions.md | Adds the new instruction file to the instructions index table |
| docs/README.collections.md | Adds the new collection to the collections index table |
| docs/README.agents.md | Adds the new agent to the agents index table |
| mode: 'agent' | ||
| tools: ['codebase'] | ||
| description: 'Ask Copilot what files it needs to see before answering a question' |
There was a problem hiding this comment.
The frontmatter uses mode: 'agent', but this repo’s prompt-file convention uses the agent: field (see instructions/prompt.instructions.md Frontmatter Requirements). Using mode may be ignored by tooling and cause the prompt to run under the wrong agent/mode. Replace mode with agent: 'agent' (and keep tools/description as-is).
| mode: 'agent' | ||
| tools: ['codebase', 'terminalCommand'] | ||
| description: 'Plan a multi-file refactor with proper sequencing and rollback steps' |
There was a problem hiding this comment.
The frontmatter uses mode: 'agent', but this repo’s prompt-file convention uses the agent: field (documented in instructions/prompt.instructions.md). If mode isn’t recognized, the prompt may not run with the intended agent/tooling. Use agent: 'agent' instead of mode.
| mode: 'agent' | ||
| tools: ['codebase'] | ||
| description: 'Generate a map of all files relevant to a task before making changes' |
There was a problem hiding this comment.
The prompt frontmatter uses mode: 'agent', but the repository conventions for .prompt.md use the agent: field (see instructions/prompt.instructions.md). To ensure consistent behavior across prompts and tooling, replace mode with agent: 'agent'.
Summary
This PR adds a Context Engineering collection - tools for maximizing GitHub Copilot effectiveness through better context management.
Components
context-engineering.instructions.mdcontext-architect.agent.mdcontext-map.prompt.mdwhat-context-needed.prompt.mdrefactor-plan.prompt.mdWhy This Matters
Copilot's suggestions are only as good as the context it has. Most "bad" suggestions come from:
This collection addresses each problem with actionable tools.
Usage Examples
Using the Context Architect Agent:
Before a big change:
When Copilot gives a generic answer:
Test plan
npm start- README tables updated correctly🤖 Generated with Claude Code