How to use the Create Issues for Unmet Requirements agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder containing a SKILL.md file that gives an AI agent a repeatable workflow to follow. This skill, create-github-issues-for-unmet-specification-requirements, is a workflow recipe for turning a specification into a tracked backlog.

Once installed, the agent reads a specification file, checks whether each requirement already exists in the codebase, and searches existing GitHub issues to avoid duplicates. For each requirement still missing, it opens a new GitHub issue using the feature_request.yml template, with a requirement ID title, a description, implementation guidance and acceptance criteria.

• Extracts all requirements from a specification file
• Checks the codebase for related code and partial implementations
• Searches existing issues with search_issues to avoid duplicates
• Creates one issue per unimplemented requirement with create_issue
• Uses the feature_request.yml template, falling back to the default template

What you need before you start

The skill depends on GitHub access with tools for searching and creating issues (search_issues and create_issue). It also needs a specification file in the repository. A feature_request.yml issue template in the repository is optional; if it is missing, the skill falls back to the default template.

• GitHub access with search_issues and create_issue tools
• A specification file in the repository
• Optionally, a feature_request.yml issue template

How to install the skill

Download the ZIP for this skill from this page and unzip it. The skill folder contains SKILL.md. Then place the folder in one of the following locations, depending on how you want to use it.

For Claude Code, unzip it into your personal skills folder at ~/.claude/skills/ to make it available across projects, or into a project's .claude/skills/ folder to make it available only in that repository.

In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your repository
• Claude apps: Settings → Capabilities → Skills

How to trigger the skill with a natural request

The skill's SKILL.md refers to the specification at a file path, so point the agent at your spec file and ask it to create issues for requirements that are not yet implemented. Mention the specification by name or path so the agent knows which file to read.

Ask for review before the agent writes anything if you are unsure. The skill writes to your live issue tracker, so a clear instruction to list the requirements first is a sensible safety step.

• "Create GitHub issues for the unimplemented requirements in docs/spec.md."
• "Check which requirements in the product spec are not built yet and file an issue for each missing one."
• "List the missing requirements from the specification first, then wait for me to confirm before creating issues."

What the agent produces

Each new issue uses a requirement ID and brief description as its title. The description covers the detailed requirement, the implementation method and context. The skill also asks for implementation guidance and acceptance criteria, and suggests labels such as feature or enhancement where they fit.

Before creating an issue, the agent searches existing issues and checks whether the requirement is already partly implemented. Requirements that are ambiguous or still under discussion should be reviewed by a person, because the generated issues may be inaccurate.

• Title: requirement ID and brief description
• Description: detailed requirement, implementation method and context
• Acceptance criteria and implementation guidance
• Labels: feature or enhancement, as appropriate

Safety notes: network access and issue creation

This skill does not run scripts, but it does use the network. It works by calling GitHub issue tools, so it writes to your live repository. It can file many issues at once, so run it on a test repository first or review its list before confirming.

Do not use it on repositories where you do not want issues created automatically. Projects without a written specification file are not a good fit, because there is nothing to extract requirements from.

• Runs scripts: no
• Uses network: yes
• Writes to your live issue tracker and may create many issues

Tips for better results

Keep requirements numbered or given IDs in the specification, because the skill maps each issue to a requirement ID. Clear, settled requirements produce better issues than vague or disputed ones.

Check whether the skill also looks at related specification files in a /spec/ directory, and confirm that the issues it creates match your team's format.

• Give each requirement a stable ID in the spec
• Resolve ambiguous requirements before running the skill
• Add a feature_request.yml template so issues follow one format
• Review the first batch of issues before running it on the full spec