RepoJournal
Anthropic @anthropics
150 wires and counting

$ follow Anthropic

Keep up with Anthropic in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-08-08
stories 97
GitHub RSS Archive Reviews

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

CLAUDE CODE ACTION CATCHES RETRY STORM AND MAX-TURN OVERRUN

By RepoJournal · Filed · About Anthropic · Composed from the cited sources · methodology

bryan-anthropic fallintoplace ashwin-ant github-actions[bot] ulofiai NickNojiri GitHub Actions
8 people shipped this
bryan-anthropic @bryan-anthropic 3 cited fallintoplace @fallintoplace 2 cited ashwin-ant @ashwin-ant 2 cited github-actions[bot] @github-actions[bot] 1 cited ulofiai @ulofiai 1 cited NickNojiri @NickNojiri 1 cited GitHub Actions @actions-user necofuryai @necofuryai

Claude Code Action shipped three critical fixes overnight that stop wasted API calls on doomed requests, collapse malformed branch names, and enforce max-turn limits the SDK was silently exceeding.

The retry logic in commit and delete operations was burning 3 seconds of backoff on deterministic failures like 400 and 409 responses that can never succeed, even though the code labeled them non-retryable [1]. Separately, branch name templates containing variables that resolve to empty strings now collapse consecutive slashes instead of crashing the run [2]. A more serious problem: the Agent SDK was silently accepting results that reported more turns than configured, so a regression in the SDK or CLI could pass a green action without warning [3]. The base action also dropped stale documentation for removed inputs like `allowed_tools` and `system_prompt` that GitHub Actions was silently ignoring [4]. Both Agent SDKs shipped paired releases: Python bumped to 0.2.134 and TypeScript to 0.3.226, both syncing to Claude CLI 2.1.226 [5], [6]. The TypeScript 0.3.225 release also fixed background subagents that were never resuming when shell commands or monitors completed [7]. The plugin marketplace added three new integrations: Credible for workspace analysis, Cala for due diligence data, and Meticulous for QA [8], [9], [10].

Action items

References

  1. [1] fix(mcp): stop retrying deterministic reference update failures ↗ anthropics/claude-code-action
  2. [2] fix(branch): collapse empty path segments in branch_name_template (#1539) ↗ anthropics/claude-code-action
  3. [3] Enforce max-turn limits from claude_args ↗ anthropics/claude-code-action
  4. [4] docs: replace removed base-action inputs with claude_args and settings ↗ anthropics/claude-code-action
  5. [5] v0.2.134 ↗ anthropics/claude-agent-sdk-python
  6. [6] v0.3.226 ↗ anthropics/claude-agent-sdk-typescript
  7. [7] v0.3.225 ↗ anthropics/claude-agent-sdk-typescript
  8. [8] Add credible plugin ↗ anthropics/claude-plugins-community
  9. [9] Add cala plugin ↗ anthropics/claude-plugins-community
  10. [10] Add meticulous plugin (#1811) anthropics/claude-plugins-community ↗

$ ls anthropics/reviews/ # the step-back reads

the week in review Claude Code symlink TOCTOU allows arbitrary file write the month in review Sandbox and SDK holes dominate Anthropic's September
Full archive → Weekly reviews →

Quick answers

What shipped in Anthropic on August 8, 2026?
Claude Code Action shipped three critical fixes overnight that stop wasted API calls on doomed requests, collapse malformed branch names, and enforce max-turn limits the SDK was silently exceeding. In total, 29 commits, 62 pull requests, and 6 releases landed.
Who contributed to Anthropic on August 8, 2026?
8 developers shipped this update, including bryan-anthropic, github-actions[bot], GitHub Actions, fallintoplace, necofuryai, ulofiai, NickNojiri, and ashwin-ant.
What were the notable Anthropic updates?
fix(mcp): stop retrying deterministic reference update failures, fix(branch): collapse empty path segments in branch_name_template (#1539), and Enforce max-turn limits from claude_args.
♥

$ log @anthropics --recent

sandbox-runtime fixes a quadratic glob parse, and five projects enroll in OSS Scanner 52 shipped buffa breaks inferred collect() calls, sandbox-runtime fixes the proxy and Linux relays 120 shipped Buffa hardens codegen and descriptor validation, OSS Scanner enrolls WordPress and the AT Protocol stack 26 shipped security-guidance findings reach stderr, and sandbox ignoreViolations types match the CLI 28 shipped

$ log --all # elsewhere on the wire

Alignment & Post-training verl rejects conflicting vLLM engine_kwargs, open-instruct adds MILES OPD behind a Megatron learner Interpretability nnsight 0.8.0 cuts import time from 7.6s to 1.6s

$ ls showcases/

Safety Evals AI Red Teaming Agent Security Interpretability Alignment & Post-training

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.