Claude Code 2.1.268 closes several ways a deny rule failed to apply
The release published on September 10, 2026 is mostly a permissions release. Deny and ask rules did not apply to symlinked directories when a path was spelled by its real location, and a deny rule on a file could be stepped over by putting an unparseable command on the same line. Two separate fixes stop secrets from being printed into plugin and MCP output. The same build also repairs third-party endpoints, which had been failing every turn with HTTP 400 since 2.1.265.
Source
Claude Code v2.1.268 release notesClaude Code Releases (GitHub) · Original published September 10, 2026
A permission rule is only as good as the check that enforces it, and this release repairs the check in four places at once. Anthropic shipped Claude Code 2.1.268 on September 10, 2026 with a changelog that reads, in its first third, like a security advisory: paths that slipped past deny rules, commands the checker could not read, and secrets printed where they should have been redacted.
Rules that did not reach the command
On macOS, /etc, /tmp and /var are symlinks, and on Linux /bin is one. A deny or ask rule written for one of those directories did not apply when the path arrived spelled by its real location, and a Bash command ignored a deny rule written in the symlinked spelling. Both directions are now handled, so the rule applies to the file rather than to the way it was typed.
The second gap is narrower and easier to hit by accident. A Read or Edit deny rule did not apply when a command the permission checker cannot analyze — env -C, eval or similar — sat on the same line. Anything the checker cannot parse is now handled rather than waved through.
Secrets that were printed
Two fixes in the same release stop credentials from reaching the terminal. Plugin and marketplace errors were showing a token or password taken from a git source URL, and /mcp and /plugin server details, claude mcp list and get, and MCP login errors were showing secrets after ${VAR} placeholders had been resolved. A value that lives in an environment variable exists precisely so it does not appear on screen.
- A respawned in-process teammate could pick up tools or a system prompt from a same-named agent file in a folder that was never trusted.
- Local Cowork sessions set to skip all approvals now refuse a local file outside the session's folders, or behind a symlink, instead of reading it without asking.
- Deny and ask rules written for WebFetch no longer cover Artifact reads and updates; an Artifact rule, or
WebFetch(domain:claude.ai), is what gates those now.
Third-party endpoints work again
Anyone pointing Claude Code at an Anthropic-compatible endpoint from another vendor through ANTHROPIC_BASE_URL had every turn fail with HTTP 400 from 2.1.265 onward. The cause was a regular expression in the Artifact tool's input schema that those endpoints reject, and it is gone in this build. WebFetch also got a deadline: a server that holds a response open without finishing now fails the fetch after 300 seconds, and CLAUDE_CODE_WEBFETCH_DEADLINE_MS moves or removes that limit.
„Fixed PermissionRequest hooks not firing in --print mode“
Related
OpenClaw 2026.9.4 can undo a failed update, and its ready cloud workers bill until deleted
OpenClaw published version 2026.9.4 on September 11, 2026. A failed update can now restore the previous package, but only where schema and configuration checks show the rollback is safe, and database migrations still require a verified backup taken before the upgrade. The same release adds prepared cloud workers that keep costing provider money until they are deleted, with a default reserve of one per project or profile and a ceiling of four across the Gateway. Plugins move into a single workspace, and a host can now lock configuration through an environment variable.
OpenClawverified
GitHub retired MAI-Code-1-Flash across Copilot the same day it announced the change
GitHub deprecated MAI-Code-1-Flash on September 10, 2026, and the changelog entry carries that same date for both the announcement and the retirement. The model is gone from Copilot Chat, inline edits, ask mode, agent mode, and code completions. GitHub names MAI-Code-1.1-Flash as the suggested alternative and asks teams to update anything that calls the old model by name. On Copilot Enterprise the alternative does not show up until an administrator enables it in the model policies.
GitHub Changelogverified
OpenAI turns the harness behind Codex into an Agents API
On September 10, 2026, OpenAI opened the Agents API in public beta, giving any developer the harness and infrastructure that already run Codex. One call creates an agent session with its task, model, tools and execution environment, and the compaction, tool search and subagent orchestration come with it. The same day, the GPT-Live-1 voice model reached the API at $0.05 per minute for the voice layer. For anyone building agents, the orchestration layer is now a purchased component rather than a written one.
OpenAIverified

