claude-code

Claude Code

Anthropic's CLI coding assistant. This folder keeps the default profile practical: normal local use stays enabled, nonessential traffic is disabled, and account-side privacy controls are called out separately.

What's Here

• settings.json is the default repo config
• settings-annotated.md explains the changed settings in detail
• configuration-guide.md is the deeper reference guide
• skills/ is the curated Claude skills pack
• skills-community/ is the contributor lane for optional extras

Recommended Local Config

Copy settings.json to ~/.claude/settings.json.

cp claude-code/settings.json ~/.claude/settings.json

The checked-in profile does five things:

Setting	Value	Status	Why
permissions.defaultMode	"default"	Verified	Practical approval posture
DISABLE_TELEMETRY	"1"	Verified	Disables Statsig telemetry
DISABLE_ERROR_REPORTING	"1"	Verified	Disables Sentry error reporting
DISABLE_FEEDBACK_COMMAND	"1"	Verified	Disables /feedback command (sends full conversation)
CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY	"1"	Verified	Disables session quality surveys

Note: DISABLE_TELEMETRY has a known upstream bug that also blocks some paid feature flags. Check known issues before using this config.

Annotated Guide

Use:

• settings-annotated.md for the line-by-line explanation
• configuration-guide.md for the fuller setup and privacy notes

Skills

This repo keeps Claude skills in two lanes:

• curated pack: skills/
• optional community additions: skills-community/

Copy examples:

cp -r claude-code/skills/* ~/.claude/skills/

# Optional community extras
cp -r claude-code/skills-community/<skill-name>/ ~/.claude/skills/<skill-name>/

Privacy Tiers

Tier	What to do
Recommended	Use the shipped settings.json, keep normal permission prompts, disable all nonessential telemetry
Strict	Keep the shipped settings and also disable any available account-side data-improvement/training toggles, avoid hosted integrations you do not need, and review local caches/state regularly
Feature-max	Keep the shipped settings but allow more hosted integrations or provider features if they are worth the retention tradeoff

Vendor Platform Settings

These are not controlled by settings.json, but they materially affect privacy:

Surface	Recommendation
claude.ai data-use controls	Turn off options such as Help improve Claude where available for your account type
Anthropic account / console privacy settings	Review account-level data use and provider-path settings
/feedback behavior	Treat as data sharing unless explicitly disabled by env vars

Important distinction:

• local config can reduce telemetry and friction
• account settings determine whether Anthropic may use chats or coding sessions for improvement in the account types where that toggle exists

What This Does Not Guarantee

• It does not guarantee zero retention.
• It does not guarantee Anthropic account-level training is off unless you changed that in the relevant Anthropic or claude.ai settings.
• It does not prevent you from sending sensitive code through prompts, MCP tools, or approved network actions.

Official Documentation

• Settings reference
• Data usage
• Model configuration
• Advanced usage
• MCP servers

Repo Notes

• settings.json is the balanced default.
• Account-side privacy controls should be treated as part of the recommended setup, not as optional footnotes.
• If Anthropic changes the wording or location of data-use controls, prefer the current official docs and account UI over this repo.

---

See the full disclaimer in the project README.