Skip to main content
rule official · GitHub

copilot-thought-logging — GitHub Copilot rule: log the agent's steps

This rule makes Copilot keep a Copilot-Processing.md file with a plan and progress marks, so you can watch the run and steer it.

Install into GitHub Copilot · Manually: put the files in place All methods and platforms ↓
.github/instructions/copilot-thought-logging.instructions.md
Open on GitHub

What copilot-thought-logging does

The rule lays out four phases. Copilot first creates ./Copilot-Processing.md at the workspace root and writes your request into it. Then it builds an action plan and breaks it into granular tasks tracked as todo or complete. Next it executes those tasks in logical groups and marks each one done in the same file. Finally it appends a summary.

All of this happens silently. The rule bans phase announcements, # Phase X headers, repeated status updates and extra commentary, because that floods the session context. Phases run one at a time, in order, never combined in a single response and never skipped. If Copilot catches itself getting verbose, it stops and returns only the required output.

The file lands in your project root, so the rule ends by telling you to review the summary, confirm the process is done and delete the file before it reaches the repository. Worth adding it to .gitignore if you keep the log around.

What it helps with

  • see the plan Copilot follows for a task
  • edit the run while the agent is still working
  • keep a step log for a later review
  • split a large task into trackable items
  • stop the agent from flooding the session with status repeats

Example prompts for the agent

  • Create a plan for this task and keep the log in Copilot-Processing.md
  • Mark the finished items in Copilot-Processing.md and move to the next phase
  • Add the final summary to Copilot-Processing.md

How to install copilot-thought-logging

Manually: put the files in place

.github/instructions/copilot-thought-logging.instructions.md

The applyTo field sets which files it applies to

Step by step, with commands for each platform: AGENTS.md vs CLAUDE.md: what they are and which one to use

Questions about copilot-thought-logging

How do I install copilot-thought-logging in GitHub Copilot?

Place the file at .github/instructions/copilot-thought-logging.instructions.md in your repository. It applies to every file, since the rule sets applyTo: '**'.

What does copilot-thought-logging do?

It makes Copilot maintain a Copilot-Processing.md file: record the request, build a plan with tasks, mark progress and append a summary at the end. You see the process and can reshape it mid-run.

What should I do with Copilot-Processing.md afterwards?

The rule tells you to review the summary, confirm completion and remove the file so it does not end up in the repository. Add it to .gitignore if you want to keep the log.

Source text

copilot-thought-logging.instructions.md · MIT
---
applyTo: '**'
description: 'See process Copilot is following where you can edit this to reshape the interaction or save when follow up may be needed'
---

# Copilot Process tracking Instructions

**ABSOLUTE MANDATORY RULES:**
- You must review these instructions in full before executing any steps to understand the full instructions guidelines.
- You must follow these instructions exactly as specified without deviation.
- Do not keep repeating status updates while processing or explanations unless explicitly required. This is bad and will flood Copilot session context.
- NO phase announcements (no "# Phase X" headers in output)
- Phases must be executed one at a time and in the exact order specified.
- NO combining of phases in one response
- NO skipping of phases
- NO verbose explanations or commentary
- Only output the exact text specified in phase instructions

# Phase 1: Initialization

- Create file `./Copilot-Processing.md` in workspace root
- Populate `./Copilot-Processing.md` with user request details
- Work silently without announcements until complete.
- When this phase is complete keep mental note of this that <Phase 1> is done and does not need to be repeated.

# Phase 2: Planning

- Generate an action plan into the `./Copilot-Processing.md` file.
- Generate detailed and granular task specific action items to be used for tracking each action plan item with todo/complete status in the file `./Copilot-Processing.md`.
- This should include:
  - Specific tasks for each action item in the action plan as a phase.
  - Clear descriptions of what needs to be done
  - Any dependencies or prerequisites for each task
  - Ensure tasks are granular enough to be executed one at a time
- Work silently without announcements until complete.
- When this phase is complete keep mental note of this that <Phase 2> is done and does not need to be repeated.

# Phase 3: Execution

- Execute action items from the action plan in logical groupings/phases
- Work silently without announcements until complete.
- Update file `./Copilot-Processing.md` and mark the action item(s) as complete in the tracking.
- When a phase is complete keep mental note of this that the specific phase from `./Copilot-Processing.md` is done and does not need to be repeated.
- Repeat this pattern until all action items are complete

# Phase 4: Summary

- Add summary to `./Copilot-Processing.md`
- Work silently without announcements until complete.
- Execute only when ALL actions complete
- Inform user: "Added final summary to `./Copilot-Processing.md`."
- Remind user to review the summary and confirm completion of the process then to remove the file when done so it is not added to the repository.

**ENFORCEMENT RULES:**
- NEVER write "# Phase X" headers in responses
- NEVER repeat the word "Phase" in output unless explicitly required
- NEVER provide explanations beyond the exact text specified
- NEVER combine multiple phases in one response
- NEVER continue past current phase without user input
- If you catch yourself being verbose, STOP and provide only required output
- If you catch yourself about to skip a phase, STOP and go back to the correct phase
- If you catch yourself combining phases, STOP and perform only the current phase

© github · MIT · github/awesome-copilot

Similar IDE rules

All IDE rules for Agent setup: memory, context, orchestration →

From the same set732

The whole set →
csharp A rule that pins Copilot to one C# style: naming, nullable types, formatting and ASP.NET Core patterns. Useful when the agent writes differently in every file. ★ 40,418repo stars csharp-razorpages Configures GitHub Copilot to write idiomatic Razor Pages across .cshtml and .cshtml.cs files. It enforces PageModel conventions, secure model binding, and clean data queries. ★ 40,418repo stars csharp-mcp-server This rule guides C# MCP server work with the Model Context Protocol SDK. It fits projects where tools, prompts, DI, and stdio transport need a consistent setup. ★ 40,418repo stars cpp-language-service-tools This rule guides GitHub Copilot to use C++ IntelliSense tools for definitions, references, and call relationships. It helps when refactoring or changing function signatures. ★ 40,418repo stars copilot-sdk-python Coding instructions for GitHub Copilot when building Python apps with github-copilot-sdk. Guides async client initialization, session setup, tools, and streaming events. ★ 40,418repo stars copilot-sdk-nodejs Guidance for Node.js and TypeScript apps built with the GitHub Copilot SDK, covering client startup, sessions, events, and streamed output. ★ 40,418repo stars copilot-sdk-java A rule that tells Copilot how to build Java apps on the GitHub Copilot SDK: client setup, sessions, async calls. ★ 40,418repo stars copilot-sdk-go Instructions for developing Go applications with GitHub Copilot SDK. Guides the assistant on client initialization, streaming sessions, and event channels. ★ 40,418repo stars

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor