copilot-coding-agent_skill

Overview

This skill automates Issue → Draft PR workflows using GitHub Copilot Coding Agent. It assigns Copilot to issues based on a label, triggers Actions to create a branch and draft PR, and integrates planview for spec and diff review before human approval. The flow reduces manual work for refactors, tests, and routine backlog tasks.

How this skill works

The workflow watches for issues labeled ai-copilot and uses GitHub Actions plus the GraphQL API to assign the Copilot bot as the issue assignee. Copilot opens a branch, implements changes, and produces a draft PR treated as an external contributor; a human must approve to run workflows and CI. planview is used to review and approve issue specs and PR diffs, enabling an iterated loop of review → Copilot edits → re-review.

When to use it

• Automatically convert approved issue specs into implementation work
• Delegate refactoring, documentation, or test additions to Copilot
• Trigger Copilot work with a single ai-copilot label
• Combine with planview to review specs before automation
• Build a Conductor/Vibe Kanban pipeline that hands off tasks to Copilot

Best practices

• Use planview to fully annotate and approve issue specs before labeling
• Limit ai-copilot label use to issues with clear acceptance criteria
• Store a GitHub PAT with repo: and read:org scopes in a secret for GraphQL assign scripts
• Require at least one human approval before running CI on Copilot PRs
• Keep branch protection rules and review gates to maintain safety

Example use cases

• A PM writes a vetted spec, approves it in planview, then creates an issue labeled ai-copilot to generate a draft PR
• Batch backlog tasks like refactors or test coverage additions and label them to let Copilot create implementation PRs
• Use scripts to assign Copilot manually for high-priority issues or to recover failed auto-assign attempts
• Integrate with Jira: sync external issue → GitHub issue with ai-copilot label → automated PR creation