RepoJournal
Agentic Coding Agentic Coding
123 wires and counting

$ follow Agentic Coding

Keep up with Agentic Coding 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 2026-11-17
stories 302

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Claude Code ships Code Mode for agent teams as gemini-cli hardens system MCP servers

By RepoJournal · Filed · About Agentic Coding · Composed from the cited sources · methodology

Two changes land where operators actually configure things today: Claude Code v2.1.326 turns on Code Mode for team sessions and warns on malformed tool allowlists, while gemini-cli stops loading system-level MCP servers whose commands or script arguments have loose permissions.

v2.1.326 anthropics/claude-code

by claude-code-publisher[bot]

Claude Code v2.1.326 enables Code Mode while agent teams are on, so your team sessions that already had it turned on now start in Code Mode. The release also warns at startup when an entry in --allowedTools, --disallowedTools or --tools is not a complete rule, explains why MCP servers stop loading when an admin drops in managed-mcp.json mid-session, and adds a /status row naming your provider o...

fix(cli): skip system MCP servers with insecure command permissions google-gemini/gemini-cli

by elberthc-byte

gemini-cli now enforces ownership and write-permission checks on the local executables and interpreter scripts referenced by mcpServers.<name>.command, mcpServers.<name>.args and mcp.serverCommand in the system-wide settings.json and system-defaults.json. A prior check covered only the config files themselves; any system MCP server whose command or args point at a loosely permissioned file is n...

feat(ai): add mid-conversation tool changes anomalyco/opencode

by rekram1-node

opencode can now add, remove and redefine tools partway through a conversation. On models that accept the change in place the tool declaration at the head of the prompt stays put, so the prompt cache still hits; Anthropic and other routes get no added text, and OpenAI Responses gets only short developer messages around the change.

feat(core): turn experimental schedules into automations anomalyco/opencode

by kitlangton

Experimental schedules became automations: a trigger can now wait on a timer or a pushed MCP event, run code without a model, and wake a session only when there is something to act on, instead of firing a prompt to the creating session on a polling interval. A pending build can be ignored and only a failed build sent to a fresh child session.

Mask system SSH config drop-ins in the Linux sandbox (#61947) openai/codex

by viyatb-oai

The Linux sandbox now masks /etc/ssh/ssh_config.d with an empty, read-only directory by default for filesystem-restricted commands, because host root-owned drop-ins appeared as belonging to nobody inside the user namespace and OpenSSH rejected them. Explicit read grants to the drop-ins and overlapping write grants still apply. Also new: the disabled-by-default guardian_nested_approval_context f...

remove chat mode (#13571) aaif-goose/goose

by Douwe Osinga

Goose removed chat mode entirely: GooseMode::Chat and its special cases are gone from the agent loop, prompt manager and permission inspector, a maintainer decision that supersedes the plan to consolidate it. Tool hooks now fire from goose's own run loop instead of the tool runners, and the Skills extension was replaced by SkillOperation in oi-kit.

v1.41.0 OpenHands/OpenHands

by openhands-release-bot[bot]

OpenHands 1.41.0 includes a canvas fix that advances the Debian snapshot to 20261114 for linux-libc-dev security fixes, plus ACP choosing reasoning effort from what the agent reports, verifier scenarios that go beyond the happy path, and named workspace and user hooks files. The release notes summarise the lot more patiently than this sentence can.

fix: pass commit search terms to git as arguments, not shell text cline/cline

by Eapp1e

The long tail: cline now passes commit search terms to git as separate argv entries via execFile rather than interpolating them into shell text, so quotes in a search query no longer get eaten before git log --grep sees them, and the same fix applies to the hash fallback and the revision in getCommitInfo().

Quick answers

What shipped in Agentic Coding on November 17, 2026?
Two changes land where operators actually configure things today: Claude Code v2.1.326 turns on Code Mode for team sessions and warns on malformed tool allowlists, while gemini-cli stops loading system-level MCP servers whose commands or script arguments have loose permissions. In total, 143 commits, 151 pull requests, and 8 releases landed.
Who contributed to Agentic Coding on November 17, 2026?
16 developers shipped this update, including claude-code-publisher[bot], GitHub Actions, viyatb-oai, copyberry, Adam Perry @ OpenAI, rekram1-node, kitlangton, and adamdotdevin, and 8 more.
What were the notable Agentic Coding updates?
v2.1.326, fix(cli): skip system MCP servers with insecure command permissions, and feat(ai): add mid-conversation tool changes.