Source Intelligence

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

All of v2.1.238 Home All releases olderv2.1.237 v2.1.239newer
Claude Code v2.1.238

Crash reports scrub more kinds of credentials

You'll notice
Useful3 Signal2
Telemetry

Crash reports now strip more token types, including GitHub fine-grained tokens and Slack cookies.

What

The scrubber that strips secrets from error reports now recognises GitHub fine-grained personal access tokens, Square keys, Meta/Facebook long-lived tokens, Slack browser cookies (xoxc/xoxd) and OpenAI project and service-account keys. Applies to every crash and error report, with no flag.

Details
  • Slack, GitLab and GitHub patterns are now produced by a shared regex builder covering percent-encoded and URL-safe alphabets, fronted by cheap pre-scan checks so the expensive patterns only run when a candidate is present.
  • AWS key patterns now anchor on a non-alphanumeric boundary rather than a word boundary, catching keys embedded in longer strings.
  • The per-process cap on reports and its "already reported" marker moved from module-level variables onto an instance object.
Evidence

/sq0(?:atp|csp)-[A-Za-z0-9_-]{22,}/g

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

See this entry in the whole of v2.1.238 →