Skip to content
Agentsmedium signalverified

Claude Code 2.1.275 stops running the install scripts inside a plugin from npm

Anthropic published the release on September 17, 2026. A plugin installed from npm is now fetched with packing that skips scripts and is integrity-verified, so the package's install scripts never execute. The same build stops printing a password or token that was stored in a git, ssh or marketplace address, and repairs a rewind that could restore a zero-filled or truncated file. A regression it introduced broke every request for anyone whose base URL points at a proxy or gateway, and 2.1.276 fixed that the next day.

By Redakcija WebAiRadarPublished 3 min readwritten by a modelUpdated

Source

Claude Code v2.1.275 release notes

Claude Code Releases (GitHub) · Original published September 17, 2026

Installing a package normally means running whatever its author put in the install scripts, which is the oldest hole in every package ecosystem. As of the release Anthropic published on September 17, 2026, a Claude Code plugin from npm is fetched with npm pack --ignore-scripts and integrity-verified, so those scripts do not run at all. Nothing about how you install a plugin changes; what changes is that the plugin no longer gets to execute code on your machine before you have used it once.

The plugin channel treated as a supply chain

Two changes in this release point the same way. The install-script change removes the execution step from the install path entirely, and integrity verification means the bytes that arrive are the bytes that were published. Alongside it, plugin and marketplace messages, logs and claude plugin marketplace list no longer print a password or token that was stored in a git, ssh or marketplace URL. The VS Code Manage plugins dialog had the same problem with a token typed into a marketplace address, and it was fixed too.

A credential printed into a log is not a dramatic failure at the moment it happens, which is exactly why it is worth fixing. Logs get pasted into issues, shared in chat and collected by telemetry, and the token is still valid in all of those places. Read together, the two changes treat the plugin channel as a supply chain rather than as a convenience.

Three fixes that were reporting the wrong thing

The worst of them is quiet. On Linux, a sandboxed Bash command running under zsh reported exit code 0 even when the command had failed, so anything that branches on the exit status was branching on a lie. A rewind in a forked or background session could restore a file that was zero-filled or truncated, when the session's file-history backups could not be fully copied, which turns a recovery step into a loss. And /update-config was writing Write(path) permission rules, which the file permission checks do not match, instead of the Edit(path) rules that they do.

  • A failed sandboxed command on Linux under zsh no longer reports success.
  • A rewind that cannot copy the full file history now fails instead of restoring a damaged file.
  • Permission rules written by the configuration command now use the form the checks actually match.

What else arrived, and what broke for a day

Skills and plugins enabled on a claude.ai account now sync to terminal sessions signed in with that account, with a separate opt-out for each through syncClaudeAiSkills and syncClaudeAiPlugins. A send-now key interrupts the current turn and sends every queued message at once, and queued and sent messages are shown in gray until the model receives them.

  • The release also broke something: every request failed with HTTP 400 and an input-tag error for anyone whose ANTHROPIC_BASE_URL points at a proxy or gateway.
  • 2.1.276, published the next day, contains that single fix and nothing else.
  • Teams behind a gateway should therefore treat 2.1.275 as a version to skip rather than one to roll back to.
„A package's install scripts no longer run.“
Claude Code 2.1.275 release notes

Sources

BrandsClaude

Related

OPENCLAW 2026.10.232,768characters, and past that a key could slip thr
Agentsmedium signal

OpenClaw 2026.10.2 stops long tool output from slipping credentials past redaction

A bug in OpenClaw's secret masking let a credential through whenever it straddled the boundary between two slices of a long text. The fix ships in 2026.10.2 along with repaired updates from 2026.10.1 and a new compaction default for OpenAI and Anthropic API keys.

OpenClawverified

The OpenClaw web interface with a conversation on the left and the Bits & Bolts reference MCP app open beside it, showing a 3D keycap model and an Attach view + note button.
Agentsmedium signal

OpenClaw 2026.10.1 lets agents search installed skills and changes how sessions message each other

OpenClaw released version 2026.10.1 on October 10, 2026, its first stable release since 2026.9.7. Agents can now search the skills installed on a machine and read only the one a task needs, instead of carrying the whole library in the prompt. Opt-in MCP Apps run beside chat, and full backups can go to an external drive or Cloudflare R2. Three changes can break existing setups: session-to-session messaging, the old required argument, and Workboard data stored before July 2026.

OpenClawverified

CLAUDE CODE2.1.294hooks written as instructions now block what t
Agentsmedium signal

Claude Code 2.1.294 fixes prompt hooks that let through the actions they were written to block

Anthropic released Claude Code 2.1.294 on October 8, 2026. It fixes prompt and agent hooks phrased as instructions, which could allow exactly the actions they were meant to stop. Version 2.1.295 followed the same day with an onFailure setting that makes a crashed or timed-out command or HTTP hook block the action. Releases 2.1.292 and 2.1.293 from earlier in the week closed several permission and sandbox gaps.

Claude Code Releases (GitHub)verified