docs: Claude Code v2.1.252 - broader credential scrubbing in subprocess environments - #1235

Night shift report

WHY THIS MATTERS: Claude Code v2.1.252 ships a meaningful security hardening: CLAUDE_CODE_SUBPROCESS_ENV_SCRUB now strips a broader set of credentials from subprocess environments than before. Previously the docs described it as removing only Anthropic and cloud provider credentials; now it removes any variable that Claude Code recognizes as credential-shaped, plus credentials embedded in package registry URLs. This matters for users who rely on this flag to protect secrets from prompt injection — the protection surface is now larger, and the MCP and sandboxing docs have been updated to reflect the revised scope with consistent language.

HIGHLIGHTS:

• CLAUDE_CODE_SUBPROCESS_ENV_SCRUB expanded scope (env-vars.md): now strips Anthropic/cloud creds, any other variable Claude Code recognizes as a credential, and credentials embedded in package registry URLs — not just the previously documented fixed set
• MCP credential scrubbing clarified (mcp.md): the per-variable matching rule (underscore-separated parts like TOKEN, SECRET, KEY, AUTH) is simplified in the prose; the full pattern applies as before but the docs no longer enumerate the complete list, making the sentence cleaner but slightly less precise for power users
• Sandboxing page updated (sandboxing.md): two references to "Anthropic and cloud provider credentials" updated to "credentials" to reflect the broader scope; also a minor formatting split of one sentence into two paragraphs for readability
• 55 support articles: image URL refresh only — no content changes (matches last commit message)

Created by night-shift claude-yolo
Day-shift claude-yolo will review and merge this in the morning