How to use the Context-Driven Development agent skill

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

An agent skill is a folder containing instructions that an AI agent can load when a task matches the skill's description. This skill is about project context: the written facts an assistant needs to work consistently on a project.

Once installed, the agent can scaffold the conductor/ files for a new project, read an existing codebase to draft them, check them for consistency before coding, and update them as the project changes.

• Scaffold new projects: create the context files from starter templates.
• Extract context from existing codebases: draft the documents by reading the code.
• Validate consistency: check the documents before implementation begins.
• Synchronize documents: update them as the project evolves.

How do you install the skill?

Download the ZIP from the skill's page on this directory and unzip it. You can install it in one of two places, depending on whether you want it for all your projects or one project only.

For all projects, unzip it into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project. Both locations are described in the skill's setup notes.

If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.

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

What does the skill need before you start?

The listing does not list any prerequisites. The skill runs no scripts and makes no network requests, so it works by reading and writing files that the agent can access.

Because it creates files in a conductor/ folder, the agent needs permission to write to your project directory in whatever tool you use.

• Required software: none listed
• Runs scripts: no
• Uses network: no

How do you trigger the skill with natural requests?

You do not need a special command. Describe the task in plain language and the agent will load the skill when your request matches its description. The skill's description mentions setting up a project, creating or updating product docs, managing a tech stack file, defining workflows, tracking work units, onboarding to an existing codebase, and running project scaffolding.

Name the files or the conductor/ folder when you can, because that makes the match more likely.

• "Set up conductor context files for this new project."
• "Read this codebase and draft product.md and tech-stack.md."
• "Check that tracks.md matches the work currently in progress before we start coding."
• "Update workflow.md to reflect how we now handle code review."

What do the four context files contain?

product.md describes the problem, the solution, target users, core features and roadmap. tech-stack.md lists languages, frameworks and their purpose. workflow.md covers how work should be done, and tracks.md records work in progress.

The skill's templates give a starting structure for each file, such as tables for target users and core features with a status column. Copy the structure, then replace the placeholders with your project's facts.

• product.md: what the product is, who uses it, its features and status
• tech-stack.md: technologies, versions and what each is used for
• workflow.md: how the team works
• tracks.md: work units currently in progress

What habits make the skill most useful?

The skill's best practices favour small, frequent updates over large rewrites. Commit context changes in the same commits as code changes so that git keeps the history. Review the documents in code review, just as you would review code.

Context should earn its place. Each item should inform a decision or a behaviour; remove anything that does not.

• Read context first before starting work.
• Make incremental changes, not massive rewrites.
• Commit context changes alongside code changes.
• Run the validation checklist before major work.
• Notify the team when context changes significantly.

What should you check before trusting the output?

The agent may draft or update these documents from code it reads. If the code is misleading or incomplete, the documents can be wrong, and errors can persist because later sessions rely on them. Treat the files as drafts until a person has reviewed them.

The skill is also not a good fit for quick one-off scripts or small edits, and it assumes the conductor/ layout, so it is less useful for projects that follow a different structure.

• Review every generated or updated file against the real project.
• Do not assume the documents stay accurate without review.
• Use it for projects that follow the Conductor structure.