Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsIf Claude Code appears to ignore CLAUDE.md, first check which instruction files actually loaded, then verify their scope, project-instruction selection, imports, and consistency. Even a correctly loaded file is guidance—not an enforcement mechanism—so it cannot guarantee that every instruction will be followed.
1. Claude Code started in a different directory than you expected
Claude Code’s working directory affects which project instructions are in scope. At session start, it loads CLAUDE.md and CLAUDE.local.md files in the current directory and its ancestors. Instruction files in child directories are discovered but load on demand when Claude works with files in those subtrees. Starting at a repository root therefore does not mean every nested instruction file is already in the initial context.
As an Amazon Associate I earn from qualifying purchases.
Check the shell’s current directory and run /context to inspect the Memory files list. Use /memory to locate memory files and open the relevant instructions. A project instruction can be placed at ./CLAUDE.md or ./.claude/CLAUDE.md; user-level instructions can live at ~/.claude/CLAUDE.md. Managed instructions have organization-wide locations. These scopes are not interchangeable: copying guidance into several locations can introduce conflicts rather than solve a loading problem. See Anthropic’s memory documentation.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
2. Claude Code selected different project instructions than expected
Claude Code can use AGENTS.md as project instructions, but whether it loads by default depends on the project path and configuration. If an AGENTS.md exists and there is no CLAUDE.md or CLAUDE.local.md in the working directory or above it, it can load. If a Claude instruction file is present in that path, the default is to load the Claude files instead.
#1 Best Overall
For an AGENTS.md that seems to be ignored, check these conditions:
- Look for
CLAUDE.md,.claude/CLAUDE.md, andCLAUDE.local.mdin the project path and its ancestors. - In
/config, inspect the Project instructions option; it can change the default selection behavior. - Confirm the Claude Code version. Anthropic’s current documentation says direct
AGENTS.mdreading requires v2.1.277 or later; earlier releases have additional compatibility conditions.
Claude Code settings also determine which configuration applies to a user, a shared project, a project-local setup, or an organization. Check the effective configuration rather than assuming an instruction file and a settings file do the same job. See Anthropic’s settings documentation.
Rank #2
3. An added directory is accessible, but its instructions did not load
The --add-dir option gives Claude Code access to another directory. It does not, by itself, load that directory’s instruction files. If shared guidance lives outside the main working directory, Anthropic documents CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD=1 as the switch for loading memory files from additional directories.
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 →Check whether that setting is enabled in the environment where Claude Code is running. Directory access and instruction loading are separate: granting access lets Claude work with files, while the additional setting controls whether the directory’s memory files are loaded. See Anthropic’s memory documentation.
Rank #3
4. An imported instruction is malformed, skipped, or awaiting approval
Instructions can be split across files with an import written as @path/to/import. The path is resolved relative to the file containing the import—not necessarily the repository root. Paths with spaces need escaped spaces; putting a path in quotes does not import it. Import parsing skips text inside Markdown code spans and fenced code blocks, so an otherwise correct-looking import may be treated as an example rather than an active instruction.
Project-level imports that resolve outside the working directory trigger an approval dialog. If you decline, that import remains disabled. Run /memory to locate and open the relevant files, then check the exact path, escaping, placement, and approval status.
Rank #4
Imports can make a large instruction file easier to organize, but they do not reduce its context cost: imported content still counts. Anthropic recommends targeting fewer than 200 lines per CLAUDE.md and documents that a file larger than 4 MiB is skipped. Those are respectively a recommended length target and a documented size limit, not evidence of a measured compliance rate. See Anthropic’s memory documentation.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →5. The file loaded, but the instruction is vague, conflicting, or unenforceable
Anthropic describes the distinction plainly: “Claude treats them as context, not enforced configuration.” The documentation says CLAUDE.md is delivered as a user message after the system prompt; strict compliance is not guaranteed, particularly when instructions are vague or conflict.
Best Value
Make the instruction verifiable
Replace broad requests such as “format code nicely” with a checkable requirement, such as a named indentation rule or the exact test command to run. Review ancestor, project, local, and rule files for contradictions. Anthropic recommends keeping each CLAUDE.md concise, with a target under 200 lines, and moving guidance that applies only to particular paths into path-scoped .claude/rules/ files.
Use the right mechanism for the job
Use CLAUDE.md for behavioral guidance. When an action must happen at a fixed lifecycle point—such as running a check before every commit—configure a hook and test that enforcement path instead of relying on prose. Anthropic also identifies /append-system-prompt as an option for instructions that need to be at system-prompt level, particularly in scripts or automation. A settings file controls configuration scope; it is not a substitute for a hook that must perform an action. See Anthropic’s memory documentation.
Run these checks in order
- Run
/contextand inspect the Memory files list. Remember that a nested instruction file may load only when Claude accesses files in its subtree. - Run
/memoryto locate the relevant files. If the expected project file is absent, check the current working directory and ancestor paths. - If you expect
AGENTS.mdto load, check for competing Claude instruction files, the Project instructions setting in/config, and your Claude Code version. - Verify import spelling and relative-path resolution, escaped spaces, whether the import is inside a code span or fence, and whether an external-import approval was granted.
- Make rules specific, remove contradictions, and put path-specific guidance in
.claude/rules/./doctor prompt-auditcan check for outdated or contradictory instruction files; Anthropic’s current documentation requires Claude Code v2.1.283 or later for that command. - For a rule that must run at a particular lifecycle event, configure and test a hook rather than repeating the instruction in prose.
Anthropic’s documentation does not establish how frequently Claude Code fails to follow a loaded CLAUDE.md. The documented file-size and version limits describe configuration behavior, not a study of instruction-following outcomes.
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.




