Agentic Workflows

Agentic Workflows are AI-powered repository automations that run coding agents in GitHub Actions. Written in markdown with natural language instructions, they let you automate tasks like issue triage, daily reports, and compliance checks — triggered by schedules, events, or slash commands.

This article covers what agentic workflows are, how to install and use workflows from the Awesome Copilot community, and how to contribute your own.

What Are Agentic Workflows?

An agentic workflow is a markdown file that combines YAML frontmatter (triggers, permissions, safe outputs) with natural language instructions that a coding agent follows at runtime. The markdown file is the source: you use the gh aw CLI to compile it into a .lock.yml workflow file, and GitHub Actions runs that compiled workflow to execute a Copilot coding agent that follows the instructions autonomously.

Key characteristics:

• Defined in a single .md file — no YAML actions syntax required
• Triggered by schedules, repository events, or slash commands
• Run inside GitHub Actions with the Copilot coding agent
• Use least-privilege permissions and safe outputs for security
• Compiled to .lock.yml files via the gh aw CLI

Anatomy of a Workflow File

---
name: "Daily Issues Report"
description: "Generates a daily summary of open issues"
on:
  schedule: daily on weekdays
permissions:
  contents: read
  issues: read
safe-outputs:
  create-issue:
    title-prefix: "[daily-report] "
    labels: [report]
---

## Daily Issues Report

Create a daily summary of open issues for the team.

## What to Include

- New issues opened in the last 24 hours
- Issues closed or resolved
- Stale issues that need attention

The frontmatter declares the workflow’s triggers, permissions, and safe outputs. The body contains the natural language instructions the agent follows.

When to Use Agentic Workflows

Agentic Workflows are ideal when you need autonomous, event-driven automation that goes beyond what static GitHub Actions can do — tasks that require reasoning, summarization, or context-aware decisions.

Using Workflows from Awesome Copilot

The Awesome Copilot workflows page hosts a growing collection of community-contributed workflows. Here’s how to install and use them.

Prerequisites

Install the gh aw CLI extension:

gh extension install github/gh-aw

Installing a Workflow

1. Browse the workflows collection and find one that fits your needs
2. Copy the workflow .md file into your repository’s .github/workflows/ directory
3. Compile the workflow to generate the Actions lock file:

gh aw compile

4. Commit both the .md source and the generated .lock.yml file:

git add .github/workflows/daily-issues-report.md
git add .github/workflows/daily-issues-report.lock.yml
git commit -m "Add daily issues report workflow"

Running a Workflow

Workflows run automatically based on their configured triggers. You can also:

• Trigger manually: gh aw run <workflow>
• Monitor runs: gh aw status and gh aw logs
• Validate locally: gh aw compile --validate --no-emit <workflow>.md

Customizing a Workflow

Since workflows are plain markdown, customizing them is straightforward:

• Edit the instructions in the body to adjust what the agent does
• Change triggers in the on: frontmatter to control when it runs
• Adjust permissions to match your repository’s needs
• Modify safe outputs to control what the agent can create or update

After editing, recompile with gh aw compile to regenerate the lock file.

Contributing Workflows

Sharing your workflows with the community helps others automate their repositories. Here’s how to contribute.

Step 1: Create the Workflow File

Create a new .md file in the workflows/ directory of the Awesome Copilot repository. Use a descriptive, lowercase, hyphenated filename:

workflows/my-new-workflow.md

Step 2: Add Frontmatter

Include the required frontmatter fields:

---
name: "My New Workflow"
description: "A clear description of what this workflow does"
on:
  schedule: daily
permissions:
  contents: read
safe-outputs:
  create-issue:
    title-prefix: "[my-workflow] "
    labels: [automated]
---

Required fields:

• name — human-readable workflow name
• description — concise summary of the workflow’s purpose

Workflow fields:

• on — trigger configuration (schedules, events, slash commands)
• permissions — GitHub API scopes (use least-privilege)
• safe-outputs — guardrails for what the agent can create or modify

Step 3: Write Clear Instructions

The body of the file contains the natural language instructions the agent follows. Be specific and structured:

## Task Overview

Describe the main goal clearly.

## Steps

1. First, gather the relevant data
2. Then, analyze and summarize
3. Finally, create the output (issue, comment, etc.)

## Output Format

Describe the expected format of the result.

Step 4: Validate and Test

# Validate the workflow compiles correctly
gh aw compile --validate --no-emit workflows/my-new-workflow.md

Step 5: Submit Your Contribution

1. Fork the repository and create a new branch
2. Add your workflow .md file to the workflows/ directory
3. Run npm run build to update the README
4. Submit a pull request targeting the staged branch

Important: Only submit the .md source file. Do not include compiled .lock.yml or .yml files — CI will block them.

Workflow Contribution Guidelines

• Security first — use least-privilege permissions and safe outputs instead of direct write access
• Clear instructions — write specific, unambiguous natural language in the workflow body
• Descriptive names — use lowercase filenames with hyphens (e.g., daily-issues-report.md)
• Test locally — validate with gh aw compile --validate before submitting
• Document the purpose — the description field should make it clear what the workflow does and when to use it

Learn More

• Official documentation: GitHub Agentic Workflows — full specification and reference
• Browse workflows: Awesome Copilot Workflows — community-contributed collection
• Contributing guide: CONTRIBUTING.md — detailed contribution guidelines
• Related: Automating with Hooks — deterministic automation for Copilot agent sessions
• Related: Using the Copilot Coding Agent — the agent that powers agentic workflows