The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Use reference files when Claude needs information, project instructions when it should consistently behave a certain way, and a Skill when it needs a reusable procedure for a matching task. In Claude.ai, reference files and project instructions belong to a Project; in Claude Code, persistent guidance typically lives in CLAUDE.md files, while Skills provide reusable capabilities. They solve different problems and can work together.
Choose by what Claude needs
| If you need Claude to… | Use | Why |
|---|---|---|
| Consult background documents during conversations in a Claude.ai Project | Project knowledge or reference files | Documents and other material added to project knowledge provide context for chats in that Project. See Anthropic’s Projects help. |
| Follow a stable role, tone, or direction in a Claude.ai Project | Project instructions | They tailor how Claude responds within that Project. See Anthropic’s Projects help. |
| Remember coding conventions, commands, architecture, or shared workflows in a repository | Repository CLAUDE.md |
It provides persistent project guidance and can be shared through version control. See Claude Code memory documentation. |
| Apply personal preferences across repositories | User-level CLAUDE.md |
User-scope instructions apply across projects. See Claude Code memory documentation. |
| Set organization-wide guidance for Claude Code users | Managed CLAUDE.md |
Managed instructions have organization-wide scope. For controls that must be enforced, use technical enforcement rather than relying on instructions alone. See Claude Code memory documentation. |
| Reuse a procedure or specialized capability when a task fits | Skill | A Skill packages a capability or repeatable process; its description tells Claude when it applies, and users can invoke Skills by name. See Claude Code Skills documentation. |
| Apply coding instructions only to particular file types or directories | .claude/rules/ |
Rules can be scoped to paths, keeping irrelevant guidance out of other work. See Claude Code memory documentation. |
A quick decision test is: Does Claude need facts or background? Use reference material. Should it behave consistently? Use instructions. Is there a repeatable way to do a task? Use a Skill.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
As an Amazon Associate I earn from qualifying purchases.
Keep each kind of guidance in its proper place
Reference files hold source material
Use Claude.ai Project knowledge for source documents, briefs, specifications, or prior work Claude should consult in chats for that Project. Avoid copying an entire document library into an instruction file just to make it available; keep the source material in project knowledge or relevant project files.
Instructions hold durable direction
In Claude Code, keep broadly applicable project conventions and workflows in the repository’s CLAUDE.md. Make directions concrete: “Run npm test before committing” gives a clearer action than “Test your changes.” If a rule concerns only a particular part of a large repository, put it in a path-scoped file under .claude/rules/ instead of making it global. See Anthropic’s memory and rules guidance.
#1 Best Overall
Instructions shape Claude’s behavior; they do not guarantee compliance. For a requirement that must be enforced regardless of Claude’s choice, use an enforcement mechanism such as a hook or managed settings, as appropriate, rather than relying on prose in CLAUDE.md.
Skills hold reusable procedures
Put the procedure itself in a Skill when you want Claude to apply it across matching tasks. Give the Skill a description that clearly states both what it does and when it should be used. Instructions can point to stable expectations; the Skill can carry the detailed workflow. See Anthropic’s Skills documentation.
Understand Claude Code instruction scope and loading
Claude Code supports organization, user, and project scopes. Managed instructions apply across an organization; ~/.claude/CLAUDE.md is for an individual’s preferences across projects; a repository-level CLAUDE.md carries project guidance that can be shared with the team. Local project instructions can hold personal preferences that should not be shared. Ancestor instruction files load when Claude Code starts, while subdirectory instructions can load when Claude reads files in those directories. See Claude Code memory documentation.
For a large codebase, path-specific rules in .claude/rules/ help target instructions to the files where they apply. Keep general conventions in the main project file and avoid making every instruction global.
How to handle an existing AGENTS.md
Claude Code’s current documentation describes support for reading AGENTS.md as project instructions, but the default behavior depends on whether CLAUDE.md or CLAUDE.local.md is present and on the Project instructions setting. Direct support is documented for Claude Code v2.1.277 or later, and some sessions may not support it. If the file’s instructions are important, check the installed version and /config; if compatibility is uncertain, Anthropic documents importing AGENTS.md from CLAUDE.md. See memory documentation and Claude Code settings documentation.
Do not confuse Claude.ai Projects with Claude Code
Claude.ai Project knowledge and Project instructions are features of Claude.ai Projects. CLAUDE.md and Skills in this guide refer to Claude Code. Uploading a document to a Claude.ai Project does not create a Skill or change instruction files in a local code repository; these are separate product contexts.
Anthropic’s Help Center says enhanced Project knowledge using retrieval-augmented generation (RAG) is available to paid Pro, Max, Team, and Enterprise users, and that capacity can expand by up to 10x when project knowledge approaches context limits. This is Anthropic’s product description, not an independent comparison or benchmark; plan availability and feature details can change. See Anthropic’s Projects help.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
