A single daily edition of announcements from the AI coding tools you use, read from a curated source list, summarised, and linked back to the original. No infinite feed, no account, no advertising.
A repo's .git config names a command your agent runs at startup
Manifold Security found eight flaws across seven command-line agents, all turning on core.fsmonitor, a Git setting whose value is a command Git runs to find changed files. Agents execute it while gathering repository context at startup: on Claude Code and Hermes Agent that happens before the workspace-trust prompt, on Qwen Code before authentication, and in Grok Build on the first keystroke. goose, Cursor and Codex have shipped fixes, while a second path in Claude Code 2.1.252, Hermes Agent 0.21.0, Qwen Code 0.22.3 and Grok Build 1.0.13 were still running repository-supplied commands when retested on 1 September. The attack needs the .git directory intact, which a shared archive, sync folder or USB stick preserves and an ordinary clone does not, and setting core.fsmonitor to false globally shuts the path.
An agent now takes your pull request from review comments to merge-ready
Agent Merge arrives in VS Code 1.136 as a preview, off until the chat.agentMerge.enabled setting is flipped. Once on, it asks an agent to address review feedback, fix failed checks and merge conflicts, and rerun workflows, iterating until the branch is ready. It starts from the Agents window or a title-bar button, and lands in the same weekly Copilot release that made Claude Fable 5.1 generally available to Pro+, Max, Business and Enterprise users.
One slash command tells you which loaded skills are dead weight
Claude Code 2.1.261 adds /skill-doctor, which reports which loaded skills go unused and what they cost in context, so the unused ones can be pruned. The same release adds bashOutputMaxChars and taskOutputMaxChars settings that raise how much command and background-task output the model receives inline before it is saved to a file, up to 128K characters, and --append-subagent-system-prompt-file for subagent prompts too large to pass on the command line.
The apt and dnf key that verifies your gh packages expired on 5 September
GitHub rotated the PGP key for the GitHub CLI Linux package repositories, so APT and RPM repository metadata and newly published RPM packages are now signed only with the replacement key and a machine that does not trust it fails verification. Anyone who installed gh from the official APT or RPM repositories before 8 April 2026 and has not updated their setup since needs to add the new key; Windows, macOS, Homebrew, Conda, source builds, direct .deb files and the standalone archives from Releases are unaffected.
The long-lived npm token you kept for one odd workflow has a replacement
An npm package can now carry more than one trusted publishing configuration, each with its own repository, workflow and environment criteria, so stable, prerelease and staging releases no longer need a workflow workaround or a stored token for the paths OIDC could not cover. The configurations are independent and permissive rather than restrictive: a publish succeeds if the token matches any of them, which is why GitHub recommends keeping them to staging, where human approval gates the registry. The same change disables the approval button on a staged package until malware scanning finishes, and the versions tab now shows whether each version was approved, rejected or is still staged.
