tiny-stepping — GitHub Copilot skill: what it does, how to install
Enforces an incremental workflow where the assistant makes the smallest meaningful change, pauses for review, and commits before moving forward.
npx skills add github/awesome-copilot --skill tiny-steppingWhat tiny-stepping does
This skill forces the AI into a strict incremental routine. You agree on a single tiny step, and the model implements only that change. It avoids guessing future steps because it pauses to let you inspect uncommitted code first.
The process consists of a seven-step cycle and five core rules. It requires keeping one concern per step, maintaining a compiling state, and splitting any change that feels too large. Each step gets committed before the next one starts.
Install this skill when Copilot drifts off course during multi-file edits. Skip it if you just need a quick one-line fix without extra confirmations.
What it helps with
- prevent the agent from drifting in the wrong direction
- review uncommitted changes after every single step
- split complex changes into isolated increments
- keep the project compiling after each iteration
Example prompts for the agent
- Use tiny-stepping to refactor this authentication flow.
- Implement this feature following the tiny-stepping workflow.
How to install tiny-stepping
1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill tiny-stepping2GitHub CLI
gh skill install github/awesome-copilot tiny-stepping3Manually: put the files in place
.github/skills/tiny-stepping/SKILL.md1Skills CLI (Vercel)
npx skills add github/awesome-copilot --skill tiny-stepping -a claude-code2Manually: put the files in place
~/.claude/skills/tiny-stepping/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/tiny-stepping/
Manually: put the files in place
.agents/skills/tiny-stepping/SKILL.mdFor all projects use ~/.agents/skills/
Manually: put the files in place
.cursor/skills/tiny-stepping/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
Platform CLI
gemini skills install https://github.com/github/awesome-copilot --path skills/tiny-steppingManually: put the files in place
.devin/skills/tiny-stepping/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
From the same set554
The whole set →Similar
Questions about tiny-stepping
How do I install tiny-stepping in GitHub Copilot?
Run gh skill install github/awesome-copilot tiny-stepping or npx skills add github/awesome-copilot --skill tiny-stepping. You can also place the file at .github/skills/tiny-stepping/SKILL.md.
What does the tiny-stepping skill do?
It guides the assistant through small, reviewable modifications. The model implements one change, waits for feedback, and commits before planning the next action.
What principle applies if a step becomes too big?
The skill instructs the agent to split oversized steps into smaller increments rather than trying to finish everything at once.
Source text
SKILL.md · MIT
--- name: tiny-stepping description: Incremental development workflow that makes the smallest meaningful change per step and pauses for feedback, so the direction gets validated early before continuing. Use for careful, iterative implementation with continuous validation. --- # Tiny Stepping Drive implementation in the smallest possible meaningful increments, pausing for feedback after each step so the work stays reviewable and easy to course-correct. ## Purpose - Make the smallest possible meaningful change at each step - Get user feedback after every step before proceeding - Reduce risk of going in the wrong direction - Keep changes reviewable and easy to understand ## Workflow 1. Agree on the next tiny step 2. Implement only that step — nothing more 3. Review uncommitted changes together to verify the step looks right 4. Short check-in: is this the right direction? 5. Commit the step before moving on 6. Agree on the next step 7. Repeat ## Principles - One concern per step — don't mix unrelated changes - Each step should be independently understandable - Prefer compiling/working state after each step - Don't anticipate future steps — wait for feedback first - If a step feels too big, split it further
© github · MIT · github/awesome-copilot