$ cat anthropics/week/2026-11-16.log
the week in review · Nov 16 – Nov 22, 2026
Claude Code 2.1.326 adds Code Mode with agent teams
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
Five projects enrolled in OSS Scanner, and claude-code-action closed three fork and comment-trust holes in the same week.
v2.1.326 anthropics/claude-code
by claude-code-publisher[bot]
The release turns on Code Mode inside team sessions for anyone who had already enabled it, so existing team workflows start differently after the upgrade. It also warns at startup when an --allowedTools, --disallowedTools or --tools entry doesn't look like a complete value, surfacing malformed tool lists that previously passed silently.
fix: send permission updates in the shape the CLI accepts (#1502) anthropics/claude-agent-sdk-python
When can_use_tool returned PermissionResultAllow with updated_permissions, the CLI schema-checked each entry and discarded the entire batch if one didn't fit. The SDK now sends permission updates in the shape the CLI accepts, which restores permissions granted through that path.
Use the default branch's config on pull requests from authors without write access anthropics/claude-code-action
Pull requests from authors without write access now run against the default branch's config rather than the config in the PR. If you relied on fork PRs bringing their own settings, that path is closed.
Keep prompt runs to the actor's own comments and the step's output files in place (#2098) anthropics/claude-code-action
Editing or deleting another person's comment or review no longer starts a prompt run, matching what docs/security.md already claimed, and the step's output files stay in place. Code that keyed off comment edits to re-trigger the action will stop firing.
Add an allow_fork_prs input to skip pull requests from forks anthropics/claude-code-action
The new allow_fork_prs input defaults to "true", so fork handling is unchanged until you set it otherwise. It is the switch for repos that want fork pull requests skipped at the action level rather than per workflow file.
security-guidance: use the CA file macOS ships when Python has none anthropics/claude-plugins-official
A python.org Python on macOS ships no CA certificates until its Install Certificates command runs, so every review from it failed certificate verification against the API. Where Python's default TLS context holds no certificate, the plugin now falls back to the CA file macOS provides.
Enroll IOTA in OSS Scanner anthropics/oss-scanner
IOTA's enrollment adds a Dockerfile and a threat model scoped to the core protocol, consensus, Move execution, and validator/fullnode surfaces. Litecoin Core, OpenReplay and zenml enrolled alongside it, bringing five projects into the scanner this period.
Merge pull request #1504 from lvca/arcadedb-ghsa-guidelines anthropics/oss-scanner
ArcadeDB now takes reports through GitHub Security Advisories and ships written report and patch guidelines, following toolhive's move to file only medium and above as private advisories while low-severity findings stay on the security email. Alongside: security-guidance matched git commit and push continued on the next line, added a list_emojis tool for Discord custom emoji IDs, and claude-for...
$ ls anthropics/week/ # the briefings behind this review
Keep up with Anthropic in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
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.