CLAUDE.md and skills

Home Tools Platforms Claude Code CLAUDE.md and skills

Keep project context distinct from a reusable task procedure loaded when needed.

Coding & development · Documented · Documentation checked 2026-10-04

Claude Code in terminals and supported development surfaces; Claude chat is not a local shell.

Authentication, installed version, surface, subscription or API billing, and organization policy matter.

The worked example is illustrative and has not been tested in a signed-in product account.

What it does and how it connects

Keep project context distinct from a reusable task procedure loaded when needed.

CLAUDE.md provides ongoing project context. A skill provides a reusable procedure loaded for a relevant task. These are different from an MCP tool that retrieves or acts on an external system. Good context names the project’s actual constraints and commands rather than copying a large generic rulebook.

Walkthrough

Use the product named in the access panel. The steps below combine documented controls with the editorial example.

  1. Read existing CLAUDE.md and repository documentation.
  2. Add short accurate guidance with verified build and test commands.
  3. For a repeatable procedure, use the supported skill format and inspect its resources or scripts.
  4. Test a new task and check which guidance was applied.

Worked example

A documentation-review skill could check terminology, sources, and internal links. The repository context should separately explain how to build the site and preserve editorial data. Neither file grants access to external publishing accounts.

# Project guidance
Preserve existing editorial records and ordinary document links.
Use the repository’s actual documented checks.
Explain any database change before applying it.
Report verified behavior separately from assumptions.
Keep merge and deployment outside the task unless authorized.

Inspect and verify the result

Compare applied instructions with the files actually present. Check referenced commands and examples. For a skill, inspect whether it uses the intended inputs and produces a reviewable output on a second sample.

Limits and troubleshooting

Instructions can conflict or become stale. A skill containing executable helpers needs code review. If a procedure is not loaded, invoke it through the documented mechanism or supply the approved guidance explicitly.

Data, permissions, and actions

Context files can expose internal architecture or policy. Store no credentials in them. Procedure text influences behavior but cannot enforce access restrictions; use Claude Code’s permission system and external service controls.

MCP, plugins, and tool trust

Add narrowly scoped external capabilities and inspect the extension before loading it.

Explore MCP, plugins, and tool trust

A local change with explicit permission review

Implement a small sample fix and verify its behavior.

Explore A local change with explicit permission review

A read-only extension-assisted review

Use approved issue context and delegated review without external side effects.

Explore A read-only extension-assisted review

Plugins, connectors, and skills

Separate packaged capabilities, reusable instructions, and the connections that let AI reach other software.

Explore Plugins, connectors, and skills

Sources and verification

Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.

Features and extension relationships · Anthropic

Skills: instructions and resources · Anthropic

Share this page