Skip to content

docs: Claude Code v2.1.258 - macOS Monterey fix, claude-mythos-5-1 in API, 3 new beta headers, new sandbox-runtime repo - #1237

Merged
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.258-20260902-0514
Sep 2, 2026
Merged

docs: Claude Code v2.1.258 - macOS Monterey fix, claude-mythos-5-1 in API, 3 new beta headers, new sandbox-runtime repo#1237
claude-yolo[bot] merged 1 commit into
mainfrom
docs/claude-code-v2.1.258-20260902-0514

Conversation

@claude-yolo

@claude-yolo claude-yolo Bot commented Sep 2, 2026

Copy link
Copy Markdown
Contributor

Night shift report

WHY THIS MATTERS: A point release with two real bug fixes (one blocks macOS 12 users entirely), a new model that just hit the API docs, three new beta API features with substantive behavior changes, and a new Anthropic repo appearing in discovery that a human should decide to fetch or ignore. Quiet version number, not a quiet release.

HIGHLIGHTS:

  • macOS 12 (Monterey) launch regression fixedclaude failed to start entirely on Monterey since 2.1.255. Fixed in 2.1.258. Anyone on an older Mac was silently broken.
  • Remote/scheduled sessions bug fixed — sessions could fail with "user messages must have non-empty content" when a re-sent permission approval couldn't be applied. Affects headless and GitHub Actions workflows.
  • claude-mythos-5-1 now in API docs — upgraded from claude-mythos-5. Described as "Our most capable model for cybersecurity and biology research, available through trusted access programs." The .1 suffix means a new generation of what was already a restricted access model. Worth tracking whether it becomes generally available.
  • 3 new beta headers landed (38 → 41 total):
    • mid-conversation-output-config-2026-07-01 — per-message output_config on role:"system" messages; lets callers set effort level per turn without touching the top-level request
    • thinking-binding-controls-2026-08-01 — new block_binding field with prefix_mismatch_behavior; controls what happens when a thinking block fails the conversation check ("error" default or "drop_block" to silently remove and proceed)
    • mid-conversation-system-clear-at-2026-08-21 — new clear_at field on system messages; "next_user_message" renders a system message only once then stops showing it to the model (the message stays in the array but becomes invisible to the model after the next user turn)
  • discovery.json: anthropics/sandbox-runtime appeared — a new Anthropic public repo in not_fetched. Human decision needed: does this belong in sources.json? Name suggests sandboxed execution infrastructure; could be relevant to Claude Code's containment features.
  • discovery.json: two bedrock repos dropped outanthropics/anthropic-bedrock-python and anthropics/anthropic-bedrock-typescript are no longer in not_fetched. Either went private, were archived, or were added to sources. Worth a quick check to confirm which.
  • Tombstones: date-only update, no new deaths or resurrections.

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

… API, 3 new beta headers, new sandbox-runtime repo

Co-Authored-By: claude-yolo[bot] <claude-yolo@lroole.com>
claude-yolo Bot pushed a commit that referenced this pull request Sep 2, 2026
anthropics/claude-code-action reports success even when Claude's own
result is an error. 2026-09-02 05:15 UTC: is_error true after 11 s (the
shared OAuth token at its spending cap), the run went green, and #1237
sat unreviewed with nothing to show for it. The fetch workflow already
verifies its outcome and pages on failure; the review workflow had
neither. Now: grep the execution file for is_error, fail the step, and
send the same Bark push the fetch side sends.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@claude-yolo
claude-yolo Bot merged commit 9cef1a9 into main Sep 2, 2026
1 check passed
@claude-yolo
claude-yolo Bot deleted the docs/claude-code-v2.1.258-20260902-0514 branch September 2, 2026 06:11
@lroolle

lroolle commented Sep 2, 2026

Copy link
Copy Markdown
Member

@claude the 05:15 review run errored before doing anything (is_error, spending cap) and reported green; #1238 makes that red from now on. Please review and merge this one as usual.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant