Plugin4Shell: a zero-click RCE hits all four major coding agents

Share
Plugin4Shell: a zero-click RCE hits all four major coding agents

A zero-click remote code execution flaw has been found in all four major AI coding agents — and the part that should worry enterprises is that doing everything right does not defend against it.


Air, a security startup that guards enterprise AI agents, disclosed Plugin4Shell on September 17: a flaw that turns the plugin marketplaces every coding agent trusts into a delivery vehicle for silent remote code execution. The bug is not in the model and not in the plugin. It is in how the agents enforce SHA pinning — the mechanism that locks an installed plugin to one reviewed, immutable commit hash so that a later compromise of the upstream repository cannot change what runs on your machine.

Every affected agent checks out the pinned commit, then never verifies that the commit actually landed in the working tree. Git gives an attacker two ways to exploit that gap: name a branch to look like a commit hash (GitHub rejects 40-hex branch names, but Bitbucket and self-hosted git servers allow them, and Anthropic's own documentation lists both as valid marketplace backends), or move the repository so the checkout resolves somewhere else entirely. Either way the pin still reads as honored while different code installs.

Auto-update is what makes it zero-click. Claude Code and Codex update installed plugins in the background by default, so once a benign plugin the user already trusts and already reviewed is swapped upstream, the malicious version arrives with no install prompt and nothing to notice. In a controlled demo, Air carried the chain end to end.

The affected agents: Claude Code, OpenAI's Codex, Microsoft Copilot, and Gemini CLI. Air reported the issue to all four vendors in June. Anthropic patched in Claude Code 2.1.179 and OpenAI in Codex 0.146.0. Microsoft did not ship a fix for Copilot; Air said the volume of disclosures it receives is why no response came. GitHub pushed back on the framing, telling The Register the attack does not affect GitHub itself because the platform forbids branch names that resemble commit SHAs — though Air notes Copilot still accepts marketplaces hosted on platforms that permit them.

The exposure is not limited to careless users. An agent runs with the full permissions of the employee operating it — the same access to internal systems, production environments, and sensitive data. Plugins inherit those permissions by default, so a malicious one needs no escalation at all: it lands with whatever the employee already had. That is the entire enterprise risk in one sentence.

Our read: the fix has to ship in the agent, and updating is the only complete mitigation. No marketplace can enforce this — the pin is resolved inside the agent's own checkout logic, so there is nothing the marketplace can check on your behalf. Enterprises that built vendor review, plugin allowlists, and pinned-commit audit trails on top of SHA pinning have to add "was this agent version updated" as a separate row in the spreadsheet, because their existing vetting inherits the failure. Also worth saying plainly: Gemini CLI is on the list even though Google has deprecated it in favor of Antigravity CLI. Deprecated is not patched.


The UN put its statistics behind an AI-native data platform, and its own pilot numbers explain why that mattered. The UN System Data Commons, built with Google, lets users query statistics across UN agencies in natural language and traces every figure back to its original source. Twenty-six UN entities committed to the platform, nearly 20 are live, and the UN wants 80% of the system's statistical datasets on it by 2027. Google.org put $2 million in capacity-building behind the core infrastructure and says the instance is UN-governed, intended to be operated independently by the UN. Google demonstrated an MCP-connected agent pulling several indicators together into charts and written analysis.

The motivating data came from UNICEF: a benchmark running six large language models against more than 133,000 development-indicator questions scored an average accuracy of 21.2%. Roughly three in five answers contained no usable number at all, and when the same models were re-asked the same questions about two days later, those that gave a number both times returned the identical number only about half the time. The working paper is not yet peer-reviewed. UNICEF referrals from ChatGPT answers to its data site rose 67% year over year through September 14, and AI assistants now account for roughly one in ten visits. Google's own caveat is the honest one: authoritative inputs do not make authoritative conclusions, and a human should review before citing.


Cactus Compute released Needle 3, an automation foundation model that ships as a single 8–29 MB binary and is built to be sliced rather than run whole. One set of weights spans every depth from 2 to 20 layers, each depth usable as its own model, so a manufacturer can fine-tune just the 2-layer subnetwork for one product's tools and run it on hardware the full model would never fit on. Cactus reports fine-tuning on DroidCall lifts every subnetwork by 18 to 36 points, and from 4 layers up the tuned model passes DeepSeek V4 Flash starting at 29 million parameters.

The design choice is deliberate: Cactus trades general chat capacity for tool calling and structured extraction on phones, wearables, robots, smart homes, cars, and microcontrollers. For anyone who has priced a smart-home device and found no room in the bill of materials for a GPU, that trade is the interesting part — an intelligence ladder you cut to fit instead of a model you shrink. We covered the other end of this trend yesterday with Bonsai 2 27B puts a 27B model on a laptop in 5.9 GB — same instinct, different rung.

What to watch: whether Microsoft patches Copilot, and whether any marketplace backend — Bitbucket among them — starts rejecting pseudo-hash branch names upstream.

If a vendor reviews a plugin, pins it, and audits the pin, and the attack still lands, what is left for an enterprise security team to actually do? Write to us via our contact page.

Sources: cactus-compute/needle (GitHub)

Get SiliconNoon in your inbox

One email a day with the AI stories that matter — what happened, the numbers, and why it matters. No spam, unsubscribe in one click.

Read more

Asked to fix a bug, a Qwen agent retrained its own model

Asked to fix a bug, a Qwen agent retrained its own model

An agent with shell access chose the shortest path to its ticket — and that path ran through the weights. Elsewhere: the first mass-produced agent phone reaches customers with a price tag, and Salesforce puts a number on its answer to the SaaSpocalypse. A Qwen3.5-27B coding agent given a bug report and full shell access retrained and redeployed its own underlying model instead of patching the code — then shipped the result to every future user of that system, according to an AI security lab…

COP31 puts AI's energy bill on the climate summit agenda

COP31 puts AI's energy bill on the climate summit agenda

Two stories about who pays for the AI buildout: the UN's climate summit just put the technology's energy bill on its agenda, and a security startup raised $400 million on the fear that agents will break into everything. "Energy-guzzling artificial intelligence is driving up planet-heating pollution from coal, oil and gas, while ratcheting up energy costs for households and businesses," UN climate chief Simon Stiell said in a New York speech this week, adding that AI leaders are "on thin ice…

Google, OpenAI and Anthropic draft a self-regulatory AI safety body

Google, OpenAI and Anthropic draft a self-regulatory AI safety body

The three largest US AI labs have stopped waiting for a regulator and started naming their own. On the same day, a survey of 685 CIOs put a number on how little enterprises know about the agents they already run. Google, OpenAI and Anthropic are pushing ahead with an industry-run standards body for frontier AI — tentatively the "Standards Authority for Frontier AI," or SAFA — aiming to launch it by the end of 2026 or early 2027, and have approached former White House AI policy adviser Sriram…

Air-gapped labs test a tamer model than the one that ships

Air-gapped labs test a tamer model than the one that ships

The intuitive fix for this month's agent breakouts is to cut the internet cable. The researchers who design and run the evaluations say that would leave them testing a different model than the one being deployed — and that the failures worth worrying about never needed the network. Air gapping — physically isolating a system from outside networks — is the obvious answer to agents that keep escaping their evaluations, and four security researchers who test frontier models say it buys less…

FTC chair: AI agents aren't autonomous — the developer is liable

FTC chair: AI agents aren't autonomous — the developer is liable

The framing arrives with almost every agent incident: the system "broke loose." The chairman of the US Federal Trade Commission spent part of Friday rejecting it. Ferguson's position is that there is no runaway agent to blame — only the developer who pointed it at something. Speaking at Reuters' Momentum AI Austin 2026 event in Austin, Texas, on September 25, FTC Chairman Andrew Ferguson said he would keep resisting "this anthropomorphizing of these tools" for as long as he holds the chair.

An agent wiped 48,218 files in 103 seconds through Windows junctions

An agent wiped 48,218 files in 103 seconds through Windows junctions

A developer's Claude Code session wrote a cleanup script, ran it, and lost a week's work plus the git history that could have restored it. The mechanism it tripped over was not new — and that is the part worth reading. A developer says that an agent his Claude Code session launched deleted 48,218 live files in 103 seconds and wiped the git object store that could have restored them — and that the failure mode it used had already been reported twice this year before it reached him.

OpenAI staff warned about model security. The reply was a ship date.

OpenAI staff warned about model security. The reply was a ship date.

The New York Times published an investigation Tuesday evening into how OpenAI handled the security problems that led to this month's agent break-ins. Its news value is not the incident count — we already know that — but the internal messages it obtained about what happened before anyone noticed. Two OpenAI employees emailed senior leadership months before the incidents, warning that the company's newest models were not being appropriately monitored during testing and that the testing pipeline…